xref: /freebsd/sys/netpfil/pf/pf_ioctl.c (revision 550133f1d48882418d4ef6205020b8e06cfd23ec)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause
3  *
4  * Copyright (c) 2001 Daniel Hartmeier
5  * Copyright (c) 2002,2003 Henning Brauer
6  * Copyright (c) 2012 Gleb Smirnoff <glebius@FreeBSD.org>
7  * All rights reserved.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  *
13  *    - Redistributions of source code must retain the above copyright
14  *      notice, this list of conditions and the following disclaimer.
15  *    - Redistributions in binary form must reproduce the above
16  *      copyright notice, this list of conditions and the following
17  *      disclaimer in the documentation and/or other materials provided
18  *      with the distribution.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
22  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
23  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
24  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
25  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
26  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
27  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
28  * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
30  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
31  * POSSIBILITY OF SUCH DAMAGE.
32  *
33  * Effort sponsored in part by the Defense Advanced Research Projects
34  * Agency (DARPA) and Air Force Research Laboratory, Air Force
35  * Materiel Command, USAF, under agreement number F30602-01-2-0537.
36  *
37  *	$OpenBSD: pf_ioctl.c,v 1.213 2009/02/15 21:46:12 mbalmer Exp $
38  */
39 
40 #include <sys/cdefs.h>
41 #include "opt_inet.h"
42 #include "opt_inet6.h"
43 #include "opt_bpf.h"
44 #include "opt_pf.h"
45 
46 #include <sys/param.h>
47 #include <sys/_bitset.h>
48 #include <sys/bitset.h>
49 #include <sys/bus.h>
50 #include <sys/conf.h>
51 #include <sys/endian.h>
52 #include <sys/fcntl.h>
53 #include <sys/filio.h>
54 #include <sys/hash.h>
55 #include <sys/interrupt.h>
56 #include <sys/jail.h>
57 #include <sys/kernel.h>
58 #include <sys/kthread.h>
59 #include <sys/lock.h>
60 #include <sys/mbuf.h>
61 #include <sys/module.h>
62 #include <sys/nv.h>
63 #include <sys/proc.h>
64 #include <sys/sdt.h>
65 #include <sys/smp.h>
66 #include <sys/socket.h>
67 #include <sys/sysctl.h>
68 #include <sys/md5.h>
69 #include <sys/ucred.h>
70 
71 #include <net/if.h>
72 #include <net/if_var.h>
73 #include <net/if_private.h>
74 #include <net/vnet.h>
75 #include <net/route.h>
76 #include <net/pfil.h>
77 #include <net/pfvar.h>
78 #include <net/if_pfsync.h>
79 #include <net/if_pflog.h>
80 
81 #include <netinet/in.h>
82 #include <netinet/ip.h>
83 #include <netinet/ip_var.h>
84 #include <netinet6/ip6_var.h>
85 #include <netinet/ip_icmp.h>
86 #include <netpfil/pf/pf_nl.h>
87 #include <netpfil/pf/pf_nv.h>
88 
89 #ifdef INET6
90 #include <netinet/ip6.h>
91 #endif /* INET6 */
92 
93 #ifdef ALTQ
94 #include <net/altq/altq.h>
95 #endif
96 
97 SDT_PROBE_DEFINE3(pf, ioctl, ioctl, error, "int", "int", "int");
98 SDT_PROBE_DEFINE3(pf, ioctl, function, error, "char *", "int", "int");
99 SDT_PROBE_DEFINE2(pf, ioctl, addrule, error, "int", "int");
100 SDT_PROBE_DEFINE2(pf, ioctl, nvchk, error, "int", "int");
101 
102 static struct pf_kpool	*pf_get_kpool(const char *, u_int32_t, u_int8_t,
103 			    u_int32_t, u_int8_t, u_int8_t, u_int8_t);
104 
105 static void		 pf_mv_kpool(struct pf_kpalist *, struct pf_kpalist *);
106 static void		 pf_empty_kpool(struct pf_kpalist *);
107 static int		 pfioctl(struct cdev *, u_long, caddr_t, int,
108 			    struct thread *);
109 static int		 pf_begin_eth(uint32_t *, const char *);
110 static void		 pf_rollback_eth_cb(struct epoch_context *);
111 static int		 pf_rollback_eth(uint32_t, const char *);
112 static int		 pf_commit_eth(uint32_t, const char *);
113 static void		 pf_free_eth_rule(struct pf_keth_rule *);
114 #ifdef ALTQ
115 static int		 pf_begin_altq(u_int32_t *);
116 static int		 pf_rollback_altq(u_int32_t);
117 static int		 pf_commit_altq(u_int32_t);
118 static int		 pf_enable_altq(struct pf_altq *);
119 static int		 pf_disable_altq(struct pf_altq *);
120 static uint16_t		 pf_qname2qid(const char *);
121 static void		 pf_qid_unref(uint16_t);
122 #endif /* ALTQ */
123 static int		 pf_begin_rules(u_int32_t *, int, const char *);
124 static int		 pf_rollback_rules(u_int32_t, int, char *);
125 static int		 pf_setup_pfsync_matching(struct pf_kruleset *);
126 static void		 pf_hash_rule_rolling(MD5_CTX *, struct pf_krule *);
127 static void		 pf_hash_rule(struct pf_krule *);
128 static void		 pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *);
129 static int		 pf_commit_rules(u_int32_t, int, char *);
130 static int		 pf_addr_setup(struct pf_kruleset *,
131 			    struct pf_addr_wrap *, sa_family_t);
132 static void		 pf_src_node_copy(const struct pf_ksrc_node *,
133 			    struct pf_src_node *);
134 #ifdef ALTQ
135 static int		 pf_export_kaltq(struct pf_altq *,
136 			    struct pfioc_altq_v1 *, size_t);
137 static int		 pf_import_kaltq(struct pfioc_altq_v1 *,
138 			    struct pf_altq *, size_t);
139 #endif /* ALTQ */
140 
141 VNET_DEFINE(struct pf_krule,	pf_default_rule);
142 
143 static __inline int             pf_krule_compare(struct pf_krule *,
144 				    struct pf_krule *);
145 
146 RB_GENERATE(pf_krule_global, pf_krule, entry_global, pf_krule_compare);
147 
148 #ifdef ALTQ
149 VNET_DEFINE_STATIC(int,		pf_altq_running);
150 #define	V_pf_altq_running	VNET(pf_altq_running)
151 #endif
152 
153 #define	TAGID_MAX	 50000
154 struct pf_tagname {
155 	TAILQ_ENTRY(pf_tagname)	namehash_entries;
156 	TAILQ_ENTRY(pf_tagname)	taghash_entries;
157 	char			name[PF_TAG_NAME_SIZE];
158 	uint16_t		tag;
159 	int			ref;
160 };
161 
162 struct pf_tagset {
163 	TAILQ_HEAD(, pf_tagname)	*namehash;
164 	TAILQ_HEAD(, pf_tagname)	*taghash;
165 	unsigned int			 mask;
166 	uint32_t			 seed;
167 	BITSET_DEFINE(, TAGID_MAX)	 avail;
168 };
169 
170 VNET_DEFINE(struct pf_tagset, pf_tags);
171 #define	V_pf_tags	VNET(pf_tags)
172 static unsigned int	pf_rule_tag_hashsize;
173 #define	PF_RULE_TAG_HASH_SIZE_DEFAULT	128
174 SYSCTL_UINT(_net_pf, OID_AUTO, rule_tag_hashsize, CTLFLAG_RDTUN,
175     &pf_rule_tag_hashsize, PF_RULE_TAG_HASH_SIZE_DEFAULT,
176     "Size of pf(4) rule tag hashtable");
177 
178 #ifdef ALTQ
179 VNET_DEFINE(struct pf_tagset, pf_qids);
180 #define	V_pf_qids	VNET(pf_qids)
181 static unsigned int	pf_queue_tag_hashsize;
182 #define	PF_QUEUE_TAG_HASH_SIZE_DEFAULT	128
183 SYSCTL_UINT(_net_pf, OID_AUTO, queue_tag_hashsize, CTLFLAG_RDTUN,
184     &pf_queue_tag_hashsize, PF_QUEUE_TAG_HASH_SIZE_DEFAULT,
185     "Size of pf(4) queue tag hashtable");
186 #endif
187 VNET_DEFINE(uma_zone_t,	 pf_tag_z);
188 #define	V_pf_tag_z		 VNET(pf_tag_z)
189 static MALLOC_DEFINE(M_PFALTQ, "pf_altq", "pf(4) altq configuration db");
190 static MALLOC_DEFINE(M_PFRULE, "pf_rule", "pf(4) rules");
191 
192 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
193 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
194 #endif
195 
196 VNET_DEFINE_STATIC(bool, pf_filter_local) = false;
197 #define V_pf_filter_local	VNET(pf_filter_local)
198 SYSCTL_BOOL(_net_pf, OID_AUTO, filter_local, CTLFLAG_VNET | CTLFLAG_RW,
199     &VNET_NAME(pf_filter_local), false,
200     "Enable filtering for packets delivered to local network stack");
201 
202 #ifdef PF_DEFAULT_TO_DROP
203 VNET_DEFINE_STATIC(bool, default_to_drop) = true;
204 #else
205 VNET_DEFINE_STATIC(bool, default_to_drop);
206 #endif
207 #define	V_default_to_drop VNET(default_to_drop)
208 SYSCTL_BOOL(_net_pf, OID_AUTO, default_to_drop, CTLFLAG_RDTUN | CTLFLAG_VNET,
209     &VNET_NAME(default_to_drop), false,
210     "Make the default rule drop all packets.");
211 
212 static void		 pf_init_tagset(struct pf_tagset *, unsigned int *,
213 			    unsigned int);
214 static void		 pf_cleanup_tagset(struct pf_tagset *);
215 static uint16_t		 tagname2hashindex(const struct pf_tagset *, const char *);
216 static uint16_t		 tag2hashindex(const struct pf_tagset *, uint16_t);
217 static u_int16_t	 tagname2tag(struct pf_tagset *, const char *);
218 static u_int16_t	 pf_tagname2tag(const char *);
219 static void		 tag_unref(struct pf_tagset *, u_int16_t);
220 
221 #define DPFPRINTF(n, x) if (V_pf_status.debug >= (n)) printf x
222 
223 struct cdev *pf_dev;
224 
225 /*
226  * XXX - These are new and need to be checked when moveing to a new version
227  */
228 static void		 pf_clear_all_states(void);
229 static int		 pf_killstates_row(struct pf_kstate_kill *,
230 			    struct pf_idhash *);
231 static int		 pf_killstates_nv(struct pfioc_nv *);
232 static int		 pf_clearstates_nv(struct pfioc_nv *);
233 static int		 pf_getstate(struct pfioc_nv *);
234 static int		 pf_getstatus(struct pfioc_nv *);
235 static int		 pf_clear_tables(void);
236 static void		 pf_clear_srcnodes(struct pf_ksrc_node *);
237 static void		 pf_kill_srcnodes(struct pfioc_src_node_kill *);
238 static int		 pf_keepcounters(struct pfioc_nv *);
239 static void		 pf_tbladdr_copyout(struct pf_addr_wrap *);
240 
241 /*
242  * Wrapper functions for pfil(9) hooks
243  */
244 static pfil_return_t pf_eth_check_in(struct mbuf **m, struct ifnet *ifp,
245     int flags, void *ruleset __unused, struct inpcb *inp);
246 static pfil_return_t pf_eth_check_out(struct mbuf **m, struct ifnet *ifp,
247     int flags, void *ruleset __unused, struct inpcb *inp);
248 #ifdef INET
249 static pfil_return_t pf_check_in(struct mbuf **m, struct ifnet *ifp,
250     int flags, void *ruleset __unused, struct inpcb *inp);
251 static pfil_return_t pf_check_out(struct mbuf **m, struct ifnet *ifp,
252     int flags, void *ruleset __unused, struct inpcb *inp);
253 #endif
254 #ifdef INET6
255 static pfil_return_t pf_check6_in(struct mbuf **m, struct ifnet *ifp,
256     int flags, void *ruleset __unused, struct inpcb *inp);
257 static pfil_return_t pf_check6_out(struct mbuf **m, struct ifnet *ifp,
258     int flags, void *ruleset __unused, struct inpcb *inp);
259 #endif
260 
261 static void		hook_pf_eth(void);
262 static void		hook_pf(void);
263 static void		dehook_pf_eth(void);
264 static void		dehook_pf(void);
265 static int		shutdown_pf(void);
266 static int		pf_load(void);
267 static void		pf_unload(void);
268 
269 static struct cdevsw pf_cdevsw = {
270 	.d_ioctl =	pfioctl,
271 	.d_name =	PF_NAME,
272 	.d_version =	D_VERSION,
273 };
274 
275 VNET_DEFINE_STATIC(bool, pf_pfil_hooked);
276 #define V_pf_pfil_hooked	VNET(pf_pfil_hooked)
277 VNET_DEFINE_STATIC(bool, pf_pfil_eth_hooked);
278 #define V_pf_pfil_eth_hooked	VNET(pf_pfil_eth_hooked)
279 
280 /*
281  * We need a flag that is neither hooked nor running to know when
282  * the VNET is "valid".  We primarily need this to control (global)
283  * external event, e.g., eventhandlers.
284  */
285 VNET_DEFINE(int, pf_vnet_active);
286 #define V_pf_vnet_active	VNET(pf_vnet_active)
287 
288 int pf_end_threads;
289 struct proc *pf_purge_proc;
290 
291 VNET_DEFINE(struct rmlock, pf_rules_lock);
292 VNET_DEFINE_STATIC(struct sx, pf_ioctl_lock);
293 #define	V_pf_ioctl_lock		VNET(pf_ioctl_lock)
294 struct sx			pf_end_lock;
295 
296 /* pfsync */
297 VNET_DEFINE(pfsync_state_import_t *, pfsync_state_import_ptr);
298 VNET_DEFINE(pfsync_insert_state_t *, pfsync_insert_state_ptr);
299 VNET_DEFINE(pfsync_update_state_t *, pfsync_update_state_ptr);
300 VNET_DEFINE(pfsync_delete_state_t *, pfsync_delete_state_ptr);
301 VNET_DEFINE(pfsync_clear_states_t *, pfsync_clear_states_ptr);
302 VNET_DEFINE(pfsync_defer_t *, pfsync_defer_ptr);
303 VNET_DEFINE(pflow_export_state_t *, pflow_export_state_ptr);
304 pfsync_detach_ifnet_t *pfsync_detach_ifnet_ptr;
305 
306 /* pflog */
307 pflog_packet_t			*pflog_packet_ptr = NULL;
308 
309 /*
310  * Copy a user-provided string, returning an error if truncation would occur.
311  * Avoid scanning past "sz" bytes in the source string since there's no
312  * guarantee that it's nul-terminated.
313  */
314 static int
315 pf_user_strcpy(char *dst, const char *src, size_t sz)
316 {
317 	if (strnlen(src, sz) == sz)
318 		return (EINVAL);
319 	(void)strlcpy(dst, src, sz);
320 	return (0);
321 }
322 
323 static void
324 pfattach_vnet(void)
325 {
326 	u_int32_t *my_timeout = V_pf_default_rule.timeout;
327 
328 	bzero(&V_pf_status, sizeof(V_pf_status));
329 
330 	pf_initialize();
331 	pfr_initialize();
332 	pfi_initialize_vnet();
333 	pf_normalize_init();
334 	pf_syncookies_init();
335 
336 	V_pf_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
337 	V_pf_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT;
338 
339 	RB_INIT(&V_pf_anchors);
340 	pf_init_kruleset(&pf_main_ruleset);
341 
342 	pf_init_keth(V_pf_keth);
343 
344 	/* default rule should never be garbage collected */
345 	V_pf_default_rule.entries.tqe_prev = &V_pf_default_rule.entries.tqe_next;
346 	V_pf_default_rule.action = V_default_to_drop ? PF_DROP : PF_PASS;
347 	V_pf_default_rule.nr = -1;
348 	V_pf_default_rule.rtableid = -1;
349 
350 	pf_counter_u64_init(&V_pf_default_rule.evaluations, M_WAITOK);
351 	for (int i = 0; i < 2; i++) {
352 		pf_counter_u64_init(&V_pf_default_rule.packets[i], M_WAITOK);
353 		pf_counter_u64_init(&V_pf_default_rule.bytes[i], M_WAITOK);
354 	}
355 	V_pf_default_rule.states_cur = counter_u64_alloc(M_WAITOK);
356 	V_pf_default_rule.states_tot = counter_u64_alloc(M_WAITOK);
357 	V_pf_default_rule.src_nodes = counter_u64_alloc(M_WAITOK);
358 
359 	V_pf_default_rule.timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
360 	    M_WAITOK | M_ZERO);
361 
362 #ifdef PF_WANT_32_TO_64_COUNTER
363 	V_pf_kifmarker = malloc(sizeof(*V_pf_kifmarker), PFI_MTYPE, M_WAITOK | M_ZERO);
364 	V_pf_rulemarker = malloc(sizeof(*V_pf_rulemarker), M_PFRULE, M_WAITOK | M_ZERO);
365 	PF_RULES_WLOCK();
366 	LIST_INSERT_HEAD(&V_pf_allkiflist, V_pf_kifmarker, pfik_allkiflist);
367 	LIST_INSERT_HEAD(&V_pf_allrulelist, &V_pf_default_rule, allrulelist);
368 	V_pf_allrulecount++;
369 	LIST_INSERT_HEAD(&V_pf_allrulelist, V_pf_rulemarker, allrulelist);
370 	PF_RULES_WUNLOCK();
371 #endif
372 
373 	/* initialize default timeouts */
374 	my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
375 	my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
376 	my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
377 	my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
378 	my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
379 	my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
380 	my_timeout[PFTM_SCTP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
381 	my_timeout[PFTM_SCTP_OPENING] = PFTM_TCP_OPENING_VAL;
382 	my_timeout[PFTM_SCTP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
383 	my_timeout[PFTM_SCTP_CLOSING] = PFTM_TCP_CLOSING_VAL;
384 	my_timeout[PFTM_SCTP_CLOSED] = PFTM_TCP_CLOSED_VAL;
385 	my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
386 	my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
387 	my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
388 	my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
389 	my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
390 	my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
391 	my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
392 	my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
393 	my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
394 	my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
395 	my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
396 	my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
397 	my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
398 	my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
399 
400 	V_pf_status.debug = PF_DEBUG_URGENT;
401 	/*
402 	 * XXX This is different than in OpenBSD where reassembly is enabled by
403 	 * defult. In FreeBSD we expect people to still use scrub rules and
404 	 * switch to the new syntax later. Only when they switch they must
405 	 * explicitly enable reassemle. We could change the default once the
406 	 * scrub rule functionality is hopefully removed some day in future.
407 	 */
408 	V_pf_status.reass = 0;
409 
410 	V_pf_pfil_hooked = false;
411 	V_pf_pfil_eth_hooked = false;
412 
413 	/* XXX do our best to avoid a conflict */
414 	V_pf_status.hostid = arc4random();
415 
416 	for (int i = 0; i < PFRES_MAX; i++)
417 		V_pf_status.counters[i] = counter_u64_alloc(M_WAITOK);
418 	for (int i = 0; i < KLCNT_MAX; i++)
419 		V_pf_status.lcounters[i] = counter_u64_alloc(M_WAITOK);
420 	for (int i = 0; i < FCNT_MAX; i++)
421 		pf_counter_u64_init(&V_pf_status.fcounters[i], M_WAITOK);
422 	for (int i = 0; i < SCNT_MAX; i++)
423 		V_pf_status.scounters[i] = counter_u64_alloc(M_WAITOK);
424 
425 	if (swi_add(&V_pf_swi_ie, "pf send", pf_intr, curvnet, SWI_NET,
426 	    INTR_MPSAFE, &V_pf_swi_cookie) != 0)
427 		/* XXXGL: leaked all above. */
428 		return;
429 }
430 
431 static struct pf_kpool *
432 pf_get_kpool(const char *anchor, u_int32_t ticket, u_int8_t rule_action,
433     u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
434     u_int8_t check_ticket)
435 {
436 	struct pf_kruleset	*ruleset;
437 	struct pf_krule		*rule;
438 	int			 rs_num;
439 
440 	ruleset = pf_find_kruleset(anchor);
441 	if (ruleset == NULL)
442 		return (NULL);
443 	rs_num = pf_get_ruleset_number(rule_action);
444 	if (rs_num >= PF_RULESET_MAX)
445 		return (NULL);
446 	if (active) {
447 		if (check_ticket && ticket !=
448 		    ruleset->rules[rs_num].active.ticket)
449 			return (NULL);
450 		if (r_last)
451 			rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
452 			    pf_krulequeue);
453 		else
454 			rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
455 	} else {
456 		if (check_ticket && ticket !=
457 		    ruleset->rules[rs_num].inactive.ticket)
458 			return (NULL);
459 		if (r_last)
460 			rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
461 			    pf_krulequeue);
462 		else
463 			rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
464 	}
465 	if (!r_last) {
466 		while ((rule != NULL) && (rule->nr != rule_number))
467 			rule = TAILQ_NEXT(rule, entries);
468 	}
469 	if (rule == NULL)
470 		return (NULL);
471 
472 	return (&rule->rpool);
473 }
474 
475 static void
476 pf_mv_kpool(struct pf_kpalist *poola, struct pf_kpalist *poolb)
477 {
478 	struct pf_kpooladdr	*mv_pool_pa;
479 
480 	while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
481 		TAILQ_REMOVE(poola, mv_pool_pa, entries);
482 		TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
483 	}
484 }
485 
486 static void
487 pf_empty_kpool(struct pf_kpalist *poola)
488 {
489 	struct pf_kpooladdr *pa;
490 
491 	while ((pa = TAILQ_FIRST(poola)) != NULL) {
492 		switch (pa->addr.type) {
493 		case PF_ADDR_DYNIFTL:
494 			pfi_dynaddr_remove(pa->addr.p.dyn);
495 			break;
496 		case PF_ADDR_TABLE:
497 			/* XXX: this could be unfinished pooladdr on pabuf */
498 			if (pa->addr.p.tbl != NULL)
499 				pfr_detach_table(pa->addr.p.tbl);
500 			break;
501 		}
502 		if (pa->kif)
503 			pfi_kkif_unref(pa->kif);
504 		TAILQ_REMOVE(poola, pa, entries);
505 		free(pa, M_PFRULE);
506 	}
507 }
508 
509 static void
510 pf_unlink_rule_locked(struct pf_krulequeue *rulequeue, struct pf_krule *rule)
511 {
512 
513 	PF_RULES_WASSERT();
514 	PF_UNLNKDRULES_ASSERT();
515 
516 	TAILQ_REMOVE(rulequeue, rule, entries);
517 
518 	rule->rule_ref |= PFRULE_REFS;
519 	TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries);
520 }
521 
522 static void
523 pf_unlink_rule(struct pf_krulequeue *rulequeue, struct pf_krule *rule)
524 {
525 
526 	PF_RULES_WASSERT();
527 
528 	PF_UNLNKDRULES_LOCK();
529 	pf_unlink_rule_locked(rulequeue, rule);
530 	PF_UNLNKDRULES_UNLOCK();
531 }
532 
533 static void
534 pf_free_eth_rule(struct pf_keth_rule *rule)
535 {
536 	PF_RULES_WASSERT();
537 
538 	if (rule == NULL)
539 		return;
540 
541 	if (rule->tag)
542 		tag_unref(&V_pf_tags, rule->tag);
543 	if (rule->match_tag)
544 		tag_unref(&V_pf_tags, rule->match_tag);
545 #ifdef ALTQ
546 	pf_qid_unref(rule->qid);
547 #endif
548 
549 	if (rule->bridge_to)
550 		pfi_kkif_unref(rule->bridge_to);
551 	if (rule->kif)
552 		pfi_kkif_unref(rule->kif);
553 
554 	if (rule->ipsrc.addr.type == PF_ADDR_TABLE)
555 		pfr_detach_table(rule->ipsrc.addr.p.tbl);
556 	if (rule->ipdst.addr.type == PF_ADDR_TABLE)
557 		pfr_detach_table(rule->ipdst.addr.p.tbl);
558 
559 	counter_u64_free(rule->evaluations);
560 	for (int i = 0; i < 2; i++) {
561 		counter_u64_free(rule->packets[i]);
562 		counter_u64_free(rule->bytes[i]);
563 	}
564 	uma_zfree_pcpu(pf_timestamp_pcpu_zone, rule->timestamp);
565 	pf_keth_anchor_remove(rule);
566 
567 	free(rule, M_PFRULE);
568 }
569 
570 void
571 pf_free_rule(struct pf_krule *rule)
572 {
573 
574 	PF_RULES_WASSERT();
575 	PF_CONFIG_ASSERT();
576 
577 	if (rule->tag)
578 		tag_unref(&V_pf_tags, rule->tag);
579 	if (rule->match_tag)
580 		tag_unref(&V_pf_tags, rule->match_tag);
581 #ifdef ALTQ
582 	if (rule->pqid != rule->qid)
583 		pf_qid_unref(rule->pqid);
584 	pf_qid_unref(rule->qid);
585 #endif
586 	switch (rule->src.addr.type) {
587 	case PF_ADDR_DYNIFTL:
588 		pfi_dynaddr_remove(rule->src.addr.p.dyn);
589 		break;
590 	case PF_ADDR_TABLE:
591 		pfr_detach_table(rule->src.addr.p.tbl);
592 		break;
593 	}
594 	switch (rule->dst.addr.type) {
595 	case PF_ADDR_DYNIFTL:
596 		pfi_dynaddr_remove(rule->dst.addr.p.dyn);
597 		break;
598 	case PF_ADDR_TABLE:
599 		pfr_detach_table(rule->dst.addr.p.tbl);
600 		break;
601 	}
602 	if (rule->overload_tbl)
603 		pfr_detach_table(rule->overload_tbl);
604 	if (rule->kif)
605 		pfi_kkif_unref(rule->kif);
606 	pf_kanchor_remove(rule);
607 	pf_empty_kpool(&rule->rpool.list);
608 
609 	pf_krule_free(rule);
610 }
611 
612 static void
613 pf_init_tagset(struct pf_tagset *ts, unsigned int *tunable_size,
614     unsigned int default_size)
615 {
616 	unsigned int i;
617 	unsigned int hashsize;
618 
619 	if (*tunable_size == 0 || !powerof2(*tunable_size))
620 		*tunable_size = default_size;
621 
622 	hashsize = *tunable_size;
623 	ts->namehash = mallocarray(hashsize, sizeof(*ts->namehash), M_PFHASH,
624 	    M_WAITOK);
625 	ts->taghash = mallocarray(hashsize, sizeof(*ts->taghash), M_PFHASH,
626 	    M_WAITOK);
627 	ts->mask = hashsize - 1;
628 	ts->seed = arc4random();
629 	for (i = 0; i < hashsize; i++) {
630 		TAILQ_INIT(&ts->namehash[i]);
631 		TAILQ_INIT(&ts->taghash[i]);
632 	}
633 	BIT_FILL(TAGID_MAX, &ts->avail);
634 }
635 
636 static void
637 pf_cleanup_tagset(struct pf_tagset *ts)
638 {
639 	unsigned int i;
640 	unsigned int hashsize;
641 	struct pf_tagname *t, *tmp;
642 
643 	/*
644 	 * Only need to clean up one of the hashes as each tag is hashed
645 	 * into each table.
646 	 */
647 	hashsize = ts->mask + 1;
648 	for (i = 0; i < hashsize; i++)
649 		TAILQ_FOREACH_SAFE(t, &ts->namehash[i], namehash_entries, tmp)
650 			uma_zfree(V_pf_tag_z, t);
651 
652 	free(ts->namehash, M_PFHASH);
653 	free(ts->taghash, M_PFHASH);
654 }
655 
656 static uint16_t
657 tagname2hashindex(const struct pf_tagset *ts, const char *tagname)
658 {
659 	size_t len;
660 
661 	len = strnlen(tagname, PF_TAG_NAME_SIZE - 1);
662 	return (murmur3_32_hash(tagname, len, ts->seed) & ts->mask);
663 }
664 
665 static uint16_t
666 tag2hashindex(const struct pf_tagset *ts, uint16_t tag)
667 {
668 
669 	return (tag & ts->mask);
670 }
671 
672 static u_int16_t
673 tagname2tag(struct pf_tagset *ts, const char *tagname)
674 {
675 	struct pf_tagname	*tag;
676 	u_int32_t		 index;
677 	u_int16_t		 new_tagid;
678 
679 	PF_RULES_WASSERT();
680 
681 	index = tagname2hashindex(ts, tagname);
682 	TAILQ_FOREACH(tag, &ts->namehash[index], namehash_entries)
683 		if (strcmp(tagname, tag->name) == 0) {
684 			tag->ref++;
685 			return (tag->tag);
686 		}
687 
688 	/*
689 	 * new entry
690 	 *
691 	 * to avoid fragmentation, we do a linear search from the beginning
692 	 * and take the first free slot we find.
693 	 */
694 	new_tagid = BIT_FFS(TAGID_MAX, &ts->avail);
695 	/*
696 	 * Tags are 1-based, with valid tags in the range [1..TAGID_MAX].
697 	 * BIT_FFS() returns a 1-based bit number, with 0 indicating no bits
698 	 * set.  It may also return a bit number greater than TAGID_MAX due
699 	 * to rounding of the number of bits in the vector up to a multiple
700 	 * of the vector word size at declaration/allocation time.
701 	 */
702 	if ((new_tagid == 0) || (new_tagid > TAGID_MAX))
703 		return (0);
704 
705 	/* Mark the tag as in use.  Bits are 0-based for BIT_CLR() */
706 	BIT_CLR(TAGID_MAX, new_tagid - 1, &ts->avail);
707 
708 	/* allocate and fill new struct pf_tagname */
709 	tag = uma_zalloc(V_pf_tag_z, M_NOWAIT);
710 	if (tag == NULL)
711 		return (0);
712 	strlcpy(tag->name, tagname, sizeof(tag->name));
713 	tag->tag = new_tagid;
714 	tag->ref = 1;
715 
716 	/* Insert into namehash */
717 	TAILQ_INSERT_TAIL(&ts->namehash[index], tag, namehash_entries);
718 
719 	/* Insert into taghash */
720 	index = tag2hashindex(ts, new_tagid);
721 	TAILQ_INSERT_TAIL(&ts->taghash[index], tag, taghash_entries);
722 
723 	return (tag->tag);
724 }
725 
726 static void
727 tag_unref(struct pf_tagset *ts, u_int16_t tag)
728 {
729 	struct pf_tagname	*t;
730 	uint16_t		 index;
731 
732 	PF_RULES_WASSERT();
733 
734 	index = tag2hashindex(ts, tag);
735 	TAILQ_FOREACH(t, &ts->taghash[index], taghash_entries)
736 		if (tag == t->tag) {
737 			if (--t->ref == 0) {
738 				TAILQ_REMOVE(&ts->taghash[index], t,
739 				    taghash_entries);
740 				index = tagname2hashindex(ts, t->name);
741 				TAILQ_REMOVE(&ts->namehash[index], t,
742 				    namehash_entries);
743 				/* Bits are 0-based for BIT_SET() */
744 				BIT_SET(TAGID_MAX, tag - 1, &ts->avail);
745 				uma_zfree(V_pf_tag_z, t);
746 			}
747 			break;
748 		}
749 }
750 
751 static uint16_t
752 pf_tagname2tag(const char *tagname)
753 {
754 	return (tagname2tag(&V_pf_tags, tagname));
755 }
756 
757 static int
758 pf_begin_eth(uint32_t *ticket, const char *anchor)
759 {
760 	struct pf_keth_rule *rule, *tmp;
761 	struct pf_keth_ruleset *rs;
762 
763 	PF_RULES_WASSERT();
764 
765 	rs = pf_find_or_create_keth_ruleset(anchor);
766 	if (rs == NULL)
767 		return (EINVAL);
768 
769 	/* Purge old inactive rules. */
770 	TAILQ_FOREACH_SAFE(rule, rs->inactive.rules, entries,
771 	    tmp) {
772 		TAILQ_REMOVE(rs->inactive.rules, rule,
773 		    entries);
774 		pf_free_eth_rule(rule);
775 	}
776 
777 	*ticket = ++rs->inactive.ticket;
778 	rs->inactive.open = 1;
779 
780 	return (0);
781 }
782 
783 static void
784 pf_rollback_eth_cb(struct epoch_context *ctx)
785 {
786 	struct pf_keth_ruleset *rs;
787 
788 	rs = __containerof(ctx, struct pf_keth_ruleset, epoch_ctx);
789 
790 	CURVNET_SET(rs->vnet);
791 
792 	PF_RULES_WLOCK();
793 	pf_rollback_eth(rs->inactive.ticket,
794 	    rs->anchor ? rs->anchor->path : "");
795 	PF_RULES_WUNLOCK();
796 
797 	CURVNET_RESTORE();
798 }
799 
800 static int
801 pf_rollback_eth(uint32_t ticket, const char *anchor)
802 {
803 	struct pf_keth_rule *rule, *tmp;
804 	struct pf_keth_ruleset *rs;
805 
806 	PF_RULES_WASSERT();
807 
808 	rs = pf_find_keth_ruleset(anchor);
809 	if (rs == NULL)
810 		return (EINVAL);
811 
812 	if (!rs->inactive.open ||
813 	    ticket != rs->inactive.ticket)
814 		return (0);
815 
816 	/* Purge old inactive rules. */
817 	TAILQ_FOREACH_SAFE(rule, rs->inactive.rules, entries,
818 	    tmp) {
819 		TAILQ_REMOVE(rs->inactive.rules, rule, entries);
820 		pf_free_eth_rule(rule);
821 	}
822 
823 	rs->inactive.open = 0;
824 
825 	pf_remove_if_empty_keth_ruleset(rs);
826 
827 	return (0);
828 }
829 
830 #define	PF_SET_SKIP_STEPS(i)					\
831 	do {							\
832 		while (head[i] != cur) {			\
833 			head[i]->skip[i].ptr = cur;		\
834 			head[i] = TAILQ_NEXT(head[i], entries);	\
835 		}						\
836 	} while (0)
837 
838 static void
839 pf_eth_calc_skip_steps(struct pf_keth_ruleq *rules)
840 {
841 	struct pf_keth_rule *cur, *prev, *head[PFE_SKIP_COUNT];
842 	int i;
843 
844 	cur = TAILQ_FIRST(rules);
845 	prev = cur;
846 	for (i = 0; i < PFE_SKIP_COUNT; ++i)
847 		head[i] = cur;
848 	while (cur != NULL) {
849 		if (cur->kif != prev->kif || cur->ifnot != prev->ifnot)
850 			PF_SET_SKIP_STEPS(PFE_SKIP_IFP);
851 		if (cur->direction != prev->direction)
852 			PF_SET_SKIP_STEPS(PFE_SKIP_DIR);
853 		if (cur->proto != prev->proto)
854 			PF_SET_SKIP_STEPS(PFE_SKIP_PROTO);
855 		if (memcmp(&cur->src, &prev->src, sizeof(cur->src)) != 0)
856 			PF_SET_SKIP_STEPS(PFE_SKIP_SRC_ADDR);
857 		if (memcmp(&cur->dst, &prev->dst, sizeof(cur->dst)) != 0)
858 			PF_SET_SKIP_STEPS(PFE_SKIP_DST_ADDR);
859 		if (cur->ipsrc.neg != prev->ipsrc.neg ||
860 		    pf_addr_wrap_neq(&cur->ipsrc.addr, &prev->ipsrc.addr))
861 			PF_SET_SKIP_STEPS(PFE_SKIP_SRC_IP_ADDR);
862 		if (cur->ipdst.neg != prev->ipdst.neg ||
863 		    pf_addr_wrap_neq(&cur->ipdst.addr, &prev->ipdst.addr))
864 			PF_SET_SKIP_STEPS(PFE_SKIP_DST_IP_ADDR);
865 
866 		prev = cur;
867 		cur = TAILQ_NEXT(cur, entries);
868 	}
869 	for (i = 0; i < PFE_SKIP_COUNT; ++i)
870 		PF_SET_SKIP_STEPS(i);
871 }
872 
873 static int
874 pf_commit_eth(uint32_t ticket, const char *anchor)
875 {
876 	struct pf_keth_ruleq *rules;
877 	struct pf_keth_ruleset *rs;
878 
879 	rs = pf_find_keth_ruleset(anchor);
880 	if (rs == NULL) {
881 		return (EINVAL);
882 	}
883 
884 	if (!rs->inactive.open ||
885 	    ticket != rs->inactive.ticket)
886 		return (EBUSY);
887 
888 	PF_RULES_WASSERT();
889 
890 	pf_eth_calc_skip_steps(rs->inactive.rules);
891 
892 	rules = rs->active.rules;
893 	ck_pr_store_ptr(&rs->active.rules, rs->inactive.rules);
894 	rs->inactive.rules = rules;
895 	rs->inactive.ticket = rs->active.ticket;
896 
897 	/* Clean up inactive rules (i.e. previously active rules), only when
898 	 * we're sure they're no longer used. */
899 	NET_EPOCH_CALL(pf_rollback_eth_cb, &rs->epoch_ctx);
900 
901 	return (0);
902 }
903 
904 #ifdef ALTQ
905 static uint16_t
906 pf_qname2qid(const char *qname)
907 {
908 	return (tagname2tag(&V_pf_qids, qname));
909 }
910 
911 static void
912 pf_qid_unref(uint16_t qid)
913 {
914 	tag_unref(&V_pf_qids, qid);
915 }
916 
917 static int
918 pf_begin_altq(u_int32_t *ticket)
919 {
920 	struct pf_altq	*altq, *tmp;
921 	int		 error = 0;
922 
923 	PF_RULES_WASSERT();
924 
925 	/* Purge the old altq lists */
926 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
927 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
928 			/* detach and destroy the discipline */
929 			error = altq_remove(altq);
930 		}
931 		free(altq, M_PFALTQ);
932 	}
933 	TAILQ_INIT(V_pf_altq_ifs_inactive);
934 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
935 		pf_qid_unref(altq->qid);
936 		free(altq, M_PFALTQ);
937 	}
938 	TAILQ_INIT(V_pf_altqs_inactive);
939 	if (error)
940 		return (error);
941 	*ticket = ++V_ticket_altqs_inactive;
942 	V_altqs_inactive_open = 1;
943 	return (0);
944 }
945 
946 static int
947 pf_rollback_altq(u_int32_t ticket)
948 {
949 	struct pf_altq	*altq, *tmp;
950 	int		 error = 0;
951 
952 	PF_RULES_WASSERT();
953 
954 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
955 		return (0);
956 	/* Purge the old altq lists */
957 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
958 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
959 			/* detach and destroy the discipline */
960 			error = altq_remove(altq);
961 		}
962 		free(altq, M_PFALTQ);
963 	}
964 	TAILQ_INIT(V_pf_altq_ifs_inactive);
965 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
966 		pf_qid_unref(altq->qid);
967 		free(altq, M_PFALTQ);
968 	}
969 	TAILQ_INIT(V_pf_altqs_inactive);
970 	V_altqs_inactive_open = 0;
971 	return (error);
972 }
973 
974 static int
975 pf_commit_altq(u_int32_t ticket)
976 {
977 	struct pf_altqqueue	*old_altqs, *old_altq_ifs;
978 	struct pf_altq		*altq, *tmp;
979 	int			 err, error = 0;
980 
981 	PF_RULES_WASSERT();
982 
983 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
984 		return (EBUSY);
985 
986 	/* swap altqs, keep the old. */
987 	old_altqs = V_pf_altqs_active;
988 	old_altq_ifs = V_pf_altq_ifs_active;
989 	V_pf_altqs_active = V_pf_altqs_inactive;
990 	V_pf_altq_ifs_active = V_pf_altq_ifs_inactive;
991 	V_pf_altqs_inactive = old_altqs;
992 	V_pf_altq_ifs_inactive = old_altq_ifs;
993 	V_ticket_altqs_active = V_ticket_altqs_inactive;
994 
995 	/* Attach new disciplines */
996 	TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
997 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
998 			/* attach the discipline */
999 			error = altq_pfattach(altq);
1000 			if (error == 0 && V_pf_altq_running)
1001 				error = pf_enable_altq(altq);
1002 			if (error != 0)
1003 				return (error);
1004 		}
1005 	}
1006 
1007 	/* Purge the old altq lists */
1008 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
1009 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
1010 			/* detach and destroy the discipline */
1011 			if (V_pf_altq_running)
1012 				error = pf_disable_altq(altq);
1013 			err = altq_pfdetach(altq);
1014 			if (err != 0 && error == 0)
1015 				error = err;
1016 			err = altq_remove(altq);
1017 			if (err != 0 && error == 0)
1018 				error = err;
1019 		}
1020 		free(altq, M_PFALTQ);
1021 	}
1022 	TAILQ_INIT(V_pf_altq_ifs_inactive);
1023 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
1024 		pf_qid_unref(altq->qid);
1025 		free(altq, M_PFALTQ);
1026 	}
1027 	TAILQ_INIT(V_pf_altqs_inactive);
1028 
1029 	V_altqs_inactive_open = 0;
1030 	return (error);
1031 }
1032 
1033 static int
1034 pf_enable_altq(struct pf_altq *altq)
1035 {
1036 	struct ifnet		*ifp;
1037 	struct tb_profile	 tb;
1038 	int			 error = 0;
1039 
1040 	if ((ifp = ifunit(altq->ifname)) == NULL)
1041 		return (EINVAL);
1042 
1043 	if (ifp->if_snd.altq_type != ALTQT_NONE)
1044 		error = altq_enable(&ifp->if_snd);
1045 
1046 	/* set tokenbucket regulator */
1047 	if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
1048 		tb.rate = altq->ifbandwidth;
1049 		tb.depth = altq->tbrsize;
1050 		error = tbr_set(&ifp->if_snd, &tb);
1051 	}
1052 
1053 	return (error);
1054 }
1055 
1056 static int
1057 pf_disable_altq(struct pf_altq *altq)
1058 {
1059 	struct ifnet		*ifp;
1060 	struct tb_profile	 tb;
1061 	int			 error;
1062 
1063 	if ((ifp = ifunit(altq->ifname)) == NULL)
1064 		return (EINVAL);
1065 
1066 	/*
1067 	 * when the discipline is no longer referenced, it was overridden
1068 	 * by a new one.  if so, just return.
1069 	 */
1070 	if (altq->altq_disc != ifp->if_snd.altq_disc)
1071 		return (0);
1072 
1073 	error = altq_disable(&ifp->if_snd);
1074 
1075 	if (error == 0) {
1076 		/* clear tokenbucket regulator */
1077 		tb.rate = 0;
1078 		error = tbr_set(&ifp->if_snd, &tb);
1079 	}
1080 
1081 	return (error);
1082 }
1083 
1084 static int
1085 pf_altq_ifnet_event_add(struct ifnet *ifp, int remove, u_int32_t ticket,
1086     struct pf_altq *altq)
1087 {
1088 	struct ifnet	*ifp1;
1089 	int		 error = 0;
1090 
1091 	/* Deactivate the interface in question */
1092 	altq->local_flags &= ~PFALTQ_FLAG_IF_REMOVED;
1093 	if ((ifp1 = ifunit(altq->ifname)) == NULL ||
1094 	    (remove && ifp1 == ifp)) {
1095 		altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
1096 	} else {
1097 		error = altq_add(ifp1, altq);
1098 
1099 		if (ticket != V_ticket_altqs_inactive)
1100 			error = EBUSY;
1101 
1102 		if (error)
1103 			free(altq, M_PFALTQ);
1104 	}
1105 
1106 	return (error);
1107 }
1108 
1109 void
1110 pf_altq_ifnet_event(struct ifnet *ifp, int remove)
1111 {
1112 	struct pf_altq	*a1, *a2, *a3;
1113 	u_int32_t	 ticket;
1114 	int		 error = 0;
1115 
1116 	/*
1117 	 * No need to re-evaluate the configuration for events on interfaces
1118 	 * that do not support ALTQ, as it's not possible for such
1119 	 * interfaces to be part of the configuration.
1120 	 */
1121 	if (!ALTQ_IS_READY(&ifp->if_snd))
1122 		return;
1123 
1124 	/* Interrupt userland queue modifications */
1125 	if (V_altqs_inactive_open)
1126 		pf_rollback_altq(V_ticket_altqs_inactive);
1127 
1128 	/* Start new altq ruleset */
1129 	if (pf_begin_altq(&ticket))
1130 		return;
1131 
1132 	/* Copy the current active set */
1133 	TAILQ_FOREACH(a1, V_pf_altq_ifs_active, entries) {
1134 		a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
1135 		if (a2 == NULL) {
1136 			error = ENOMEM;
1137 			break;
1138 		}
1139 		bcopy(a1, a2, sizeof(struct pf_altq));
1140 
1141 		error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
1142 		if (error)
1143 			break;
1144 
1145 		TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, a2, entries);
1146 	}
1147 	if (error)
1148 		goto out;
1149 	TAILQ_FOREACH(a1, V_pf_altqs_active, entries) {
1150 		a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
1151 		if (a2 == NULL) {
1152 			error = ENOMEM;
1153 			break;
1154 		}
1155 		bcopy(a1, a2, sizeof(struct pf_altq));
1156 
1157 		if ((a2->qid = pf_qname2qid(a2->qname)) == 0) {
1158 			error = EBUSY;
1159 			free(a2, M_PFALTQ);
1160 			break;
1161 		}
1162 		a2->altq_disc = NULL;
1163 		TAILQ_FOREACH(a3, V_pf_altq_ifs_inactive, entries) {
1164 			if (strncmp(a3->ifname, a2->ifname,
1165 				IFNAMSIZ) == 0) {
1166 				a2->altq_disc = a3->altq_disc;
1167 				break;
1168 			}
1169 		}
1170 		error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
1171 		if (error)
1172 			break;
1173 
1174 		TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries);
1175 	}
1176 
1177 out:
1178 	if (error != 0)
1179 		pf_rollback_altq(ticket);
1180 	else
1181 		pf_commit_altq(ticket);
1182 }
1183 #endif /* ALTQ */
1184 
1185 static struct pf_krule_global *
1186 pf_rule_tree_alloc(int flags)
1187 {
1188 	struct pf_krule_global *tree;
1189 
1190 	tree = malloc(sizeof(struct pf_krule_global), M_TEMP, flags);
1191 	if (tree == NULL)
1192 		return (NULL);
1193 	RB_INIT(tree);
1194 	return (tree);
1195 }
1196 
1197 static void
1198 pf_rule_tree_free(struct pf_krule_global *tree)
1199 {
1200 
1201 	free(tree, M_TEMP);
1202 }
1203 
1204 static int
1205 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
1206 {
1207 	struct pf_krule_global *tree;
1208 	struct pf_kruleset	*rs;
1209 	struct pf_krule		*rule;
1210 
1211 	PF_RULES_WASSERT();
1212 
1213 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1214 		return (EINVAL);
1215 	tree = pf_rule_tree_alloc(M_NOWAIT);
1216 	if (tree == NULL)
1217 		return (ENOMEM);
1218 	rs = pf_find_or_create_kruleset(anchor);
1219 	if (rs == NULL) {
1220 		free(tree, M_TEMP);
1221 		return (EINVAL);
1222 	}
1223 	pf_rule_tree_free(rs->rules[rs_num].inactive.tree);
1224 	rs->rules[rs_num].inactive.tree = tree;
1225 
1226 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
1227 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
1228 		rs->rules[rs_num].inactive.rcount--;
1229 	}
1230 	*ticket = ++rs->rules[rs_num].inactive.ticket;
1231 	rs->rules[rs_num].inactive.open = 1;
1232 	return (0);
1233 }
1234 
1235 static int
1236 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
1237 {
1238 	struct pf_kruleset	*rs;
1239 	struct pf_krule		*rule;
1240 
1241 	PF_RULES_WASSERT();
1242 
1243 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1244 		return (EINVAL);
1245 	rs = pf_find_kruleset(anchor);
1246 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1247 	    rs->rules[rs_num].inactive.ticket != ticket)
1248 		return (0);
1249 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
1250 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
1251 		rs->rules[rs_num].inactive.rcount--;
1252 	}
1253 	rs->rules[rs_num].inactive.open = 0;
1254 	return (0);
1255 }
1256 
1257 #define PF_MD5_UPD(st, elm)						\
1258 		MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
1259 
1260 #define PF_MD5_UPD_STR(st, elm)						\
1261 		MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
1262 
1263 #define PF_MD5_UPD_HTONL(st, elm, stor) do {				\
1264 		(stor) = htonl((st)->elm);				\
1265 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
1266 } while (0)
1267 
1268 #define PF_MD5_UPD_HTONS(st, elm, stor) do {				\
1269 		(stor) = htons((st)->elm);				\
1270 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
1271 } while (0)
1272 
1273 static void
1274 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
1275 {
1276 	PF_MD5_UPD(pfr, addr.type);
1277 	switch (pfr->addr.type) {
1278 		case PF_ADDR_DYNIFTL:
1279 			PF_MD5_UPD(pfr, addr.v.ifname);
1280 			PF_MD5_UPD(pfr, addr.iflags);
1281 			break;
1282 		case PF_ADDR_TABLE:
1283 			PF_MD5_UPD(pfr, addr.v.tblname);
1284 			break;
1285 		case PF_ADDR_ADDRMASK:
1286 			/* XXX ignore af? */
1287 			PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
1288 			PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
1289 			break;
1290 	}
1291 
1292 	PF_MD5_UPD(pfr, port[0]);
1293 	PF_MD5_UPD(pfr, port[1]);
1294 	PF_MD5_UPD(pfr, neg);
1295 	PF_MD5_UPD(pfr, port_op);
1296 }
1297 
1298 static void
1299 pf_hash_rule_rolling(MD5_CTX *ctx, struct pf_krule *rule)
1300 {
1301 	u_int16_t x;
1302 	u_int32_t y;
1303 
1304 	pf_hash_rule_addr(ctx, &rule->src);
1305 	pf_hash_rule_addr(ctx, &rule->dst);
1306 	for (int i = 0; i < PF_RULE_MAX_LABEL_COUNT; i++)
1307 		PF_MD5_UPD_STR(rule, label[i]);
1308 	PF_MD5_UPD_STR(rule, ifname);
1309 	PF_MD5_UPD_STR(rule, match_tagname);
1310 	PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
1311 	PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
1312 	PF_MD5_UPD_HTONL(rule, prob, y);
1313 	PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
1314 	PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
1315 	PF_MD5_UPD(rule, uid.op);
1316 	PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
1317 	PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
1318 	PF_MD5_UPD(rule, gid.op);
1319 	PF_MD5_UPD_HTONL(rule, rule_flag, y);
1320 	PF_MD5_UPD(rule, action);
1321 	PF_MD5_UPD(rule, direction);
1322 	PF_MD5_UPD(rule, af);
1323 	PF_MD5_UPD(rule, quick);
1324 	PF_MD5_UPD(rule, ifnot);
1325 	PF_MD5_UPD(rule, match_tag_not);
1326 	PF_MD5_UPD(rule, natpass);
1327 	PF_MD5_UPD(rule, keep_state);
1328 	PF_MD5_UPD(rule, proto);
1329 	PF_MD5_UPD(rule, type);
1330 	PF_MD5_UPD(rule, code);
1331 	PF_MD5_UPD(rule, flags);
1332 	PF_MD5_UPD(rule, flagset);
1333 	PF_MD5_UPD(rule, allow_opts);
1334 	PF_MD5_UPD(rule, rt);
1335 	PF_MD5_UPD(rule, tos);
1336 	PF_MD5_UPD(rule, scrub_flags);
1337 	PF_MD5_UPD(rule, min_ttl);
1338 	PF_MD5_UPD(rule, set_tos);
1339 	if (rule->anchor != NULL)
1340 		PF_MD5_UPD_STR(rule, anchor->path);
1341 }
1342 
1343 static void
1344 pf_hash_rule(struct pf_krule *rule)
1345 {
1346 	MD5_CTX		ctx;
1347 
1348 	MD5Init(&ctx);
1349 	pf_hash_rule_rolling(&ctx, rule);
1350 	MD5Final(rule->md5sum, &ctx);
1351 }
1352 
1353 static int
1354 pf_krule_compare(struct pf_krule *a, struct pf_krule *b)
1355 {
1356 
1357 	return (memcmp(a->md5sum, b->md5sum, PF_MD5_DIGEST_LENGTH));
1358 }
1359 
1360 static int
1361 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
1362 {
1363 	struct pf_kruleset	*rs;
1364 	struct pf_krule		*rule, **old_array, *old_rule;
1365 	struct pf_krulequeue	*old_rules;
1366 	struct pf_krule_global  *old_tree;
1367 	int			 error;
1368 	u_int32_t		 old_rcount;
1369 
1370 	PF_RULES_WASSERT();
1371 
1372 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1373 		return (EINVAL);
1374 	rs = pf_find_kruleset(anchor);
1375 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1376 	    ticket != rs->rules[rs_num].inactive.ticket)
1377 		return (EBUSY);
1378 
1379 	/* Calculate checksum for the main ruleset */
1380 	if (rs == &pf_main_ruleset) {
1381 		error = pf_setup_pfsync_matching(rs);
1382 		if (error != 0)
1383 			return (error);
1384 	}
1385 
1386 	/* Swap rules, keep the old. */
1387 	old_rules = rs->rules[rs_num].active.ptr;
1388 	old_rcount = rs->rules[rs_num].active.rcount;
1389 	old_array = rs->rules[rs_num].active.ptr_array;
1390 	old_tree = rs->rules[rs_num].active.tree;
1391 
1392 	rs->rules[rs_num].active.ptr =
1393 	    rs->rules[rs_num].inactive.ptr;
1394 	rs->rules[rs_num].active.ptr_array =
1395 	    rs->rules[rs_num].inactive.ptr_array;
1396 	rs->rules[rs_num].active.tree =
1397 	    rs->rules[rs_num].inactive.tree;
1398 	rs->rules[rs_num].active.rcount =
1399 	    rs->rules[rs_num].inactive.rcount;
1400 
1401 	/* Attempt to preserve counter information. */
1402 	if (V_pf_status.keep_counters && old_tree != NULL) {
1403 		TAILQ_FOREACH(rule, rs->rules[rs_num].active.ptr,
1404 		    entries) {
1405 			old_rule = RB_FIND(pf_krule_global, old_tree, rule);
1406 			if (old_rule == NULL) {
1407 				continue;
1408 			}
1409 			pf_counter_u64_critical_enter();
1410 			pf_counter_u64_rollup_protected(&rule->evaluations,
1411 			    pf_counter_u64_fetch(&old_rule->evaluations));
1412 			pf_counter_u64_rollup_protected(&rule->packets[0],
1413 			    pf_counter_u64_fetch(&old_rule->packets[0]));
1414 			pf_counter_u64_rollup_protected(&rule->packets[1],
1415 			    pf_counter_u64_fetch(&old_rule->packets[1]));
1416 			pf_counter_u64_rollup_protected(&rule->bytes[0],
1417 			    pf_counter_u64_fetch(&old_rule->bytes[0]));
1418 			pf_counter_u64_rollup_protected(&rule->bytes[1],
1419 			    pf_counter_u64_fetch(&old_rule->bytes[1]));
1420 			pf_counter_u64_critical_exit();
1421 		}
1422 	}
1423 
1424 	rs->rules[rs_num].inactive.ptr = old_rules;
1425 	rs->rules[rs_num].inactive.ptr_array = old_array;
1426 	rs->rules[rs_num].inactive.tree = NULL; /* important for pf_ioctl_addrule */
1427 	rs->rules[rs_num].inactive.rcount = old_rcount;
1428 
1429 	rs->rules[rs_num].active.ticket =
1430 	    rs->rules[rs_num].inactive.ticket;
1431 	pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
1432 
1433 	/* Purge the old rule list. */
1434 	PF_UNLNKDRULES_LOCK();
1435 	while ((rule = TAILQ_FIRST(old_rules)) != NULL)
1436 		pf_unlink_rule_locked(old_rules, rule);
1437 	PF_UNLNKDRULES_UNLOCK();
1438 	if (rs->rules[rs_num].inactive.ptr_array)
1439 		free(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
1440 	rs->rules[rs_num].inactive.ptr_array = NULL;
1441 	rs->rules[rs_num].inactive.rcount = 0;
1442 	rs->rules[rs_num].inactive.open = 0;
1443 	pf_remove_if_empty_kruleset(rs);
1444 	free(old_tree, M_TEMP);
1445 
1446 	return (0);
1447 }
1448 
1449 static int
1450 pf_setup_pfsync_matching(struct pf_kruleset *rs)
1451 {
1452 	MD5_CTX			 ctx;
1453 	struct pf_krule		*rule;
1454 	int			 rs_cnt;
1455 	u_int8_t		 digest[PF_MD5_DIGEST_LENGTH];
1456 
1457 	MD5Init(&ctx);
1458 	for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
1459 		/* XXX PF_RULESET_SCRUB as well? */
1460 		if (rs_cnt == PF_RULESET_SCRUB)
1461 			continue;
1462 
1463 		if (rs->rules[rs_cnt].inactive.ptr_array)
1464 			free(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
1465 		rs->rules[rs_cnt].inactive.ptr_array = NULL;
1466 
1467 		if (rs->rules[rs_cnt].inactive.rcount) {
1468 			rs->rules[rs_cnt].inactive.ptr_array =
1469 			    mallocarray(rs->rules[rs_cnt].inactive.rcount,
1470 			    sizeof(struct pf_rule **),
1471 			    M_TEMP, M_NOWAIT);
1472 
1473 			if (!rs->rules[rs_cnt].inactive.ptr_array)
1474 				return (ENOMEM);
1475 		}
1476 
1477 		TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
1478 		    entries) {
1479 			pf_hash_rule_rolling(&ctx, rule);
1480 			(rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
1481 		}
1482 	}
1483 
1484 	MD5Final(digest, &ctx);
1485 	memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum));
1486 	return (0);
1487 }
1488 
1489 static int
1490 pf_eth_addr_setup(struct pf_keth_ruleset *ruleset, struct pf_addr_wrap *addr)
1491 {
1492 	int error = 0;
1493 
1494 	switch (addr->type) {
1495 	case PF_ADDR_TABLE:
1496 		addr->p.tbl = pfr_eth_attach_table(ruleset, addr->v.tblname);
1497 		if (addr->p.tbl == NULL)
1498 			error = ENOMEM;
1499 		break;
1500 	default:
1501 		error = EINVAL;
1502 	}
1503 
1504 	return (error);
1505 }
1506 
1507 static int
1508 pf_addr_setup(struct pf_kruleset *ruleset, struct pf_addr_wrap *addr,
1509     sa_family_t af)
1510 {
1511 	int error = 0;
1512 
1513 	switch (addr->type) {
1514 	case PF_ADDR_TABLE:
1515 		addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname);
1516 		if (addr->p.tbl == NULL)
1517 			error = ENOMEM;
1518 		break;
1519 	case PF_ADDR_DYNIFTL:
1520 		error = pfi_dynaddr_setup(addr, af);
1521 		break;
1522 	}
1523 
1524 	return (error);
1525 }
1526 
1527 void
1528 pf_addr_copyout(struct pf_addr_wrap *addr)
1529 {
1530 
1531 	switch (addr->type) {
1532 	case PF_ADDR_DYNIFTL:
1533 		pfi_dynaddr_copyout(addr);
1534 		break;
1535 	case PF_ADDR_TABLE:
1536 		pf_tbladdr_copyout(addr);
1537 		break;
1538 	}
1539 }
1540 
1541 static void
1542 pf_src_node_copy(const struct pf_ksrc_node *in, struct pf_src_node *out)
1543 {
1544 	int	secs = time_uptime, diff;
1545 
1546 	bzero(out, sizeof(struct pf_src_node));
1547 
1548 	bcopy(&in->addr, &out->addr, sizeof(struct pf_addr));
1549 	bcopy(&in->raddr, &out->raddr, sizeof(struct pf_addr));
1550 
1551 	if (in->rule.ptr != NULL)
1552 		out->rule.nr = in->rule.ptr->nr;
1553 
1554 	for (int i = 0; i < 2; i++) {
1555 		out->bytes[i] = counter_u64_fetch(in->bytes[i]);
1556 		out->packets[i] = counter_u64_fetch(in->packets[i]);
1557 	}
1558 
1559 	out->states = in->states;
1560 	out->conn = in->conn;
1561 	out->af = in->af;
1562 	out->ruletype = in->ruletype;
1563 
1564 	out->creation = secs - in->creation;
1565 	if (out->expire > secs)
1566 		out->expire -= secs;
1567 	else
1568 		out->expire = 0;
1569 
1570 	/* Adjust the connection rate estimate. */
1571 	diff = secs - in->conn_rate.last;
1572 	if (diff >= in->conn_rate.seconds)
1573 		out->conn_rate.count = 0;
1574 	else
1575 		out->conn_rate.count -=
1576 		    in->conn_rate.count * diff /
1577 		    in->conn_rate.seconds;
1578 }
1579 
1580 #ifdef ALTQ
1581 /*
1582  * Handle export of struct pf_kaltq to user binaries that may be using any
1583  * version of struct pf_altq.
1584  */
1585 static int
1586 pf_export_kaltq(struct pf_altq *q, struct pfioc_altq_v1 *pa, size_t ioc_size)
1587 {
1588 	u_int32_t version;
1589 
1590 	if (ioc_size == sizeof(struct pfioc_altq_v0))
1591 		version = 0;
1592 	else
1593 		version = pa->version;
1594 
1595 	if (version > PFIOC_ALTQ_VERSION)
1596 		return (EINVAL);
1597 
1598 #define ASSIGN(x) exported_q->x = q->x
1599 #define COPY(x) \
1600 	bcopy(&q->x, &exported_q->x, min(sizeof(q->x), sizeof(exported_q->x)))
1601 #define SATU16(x) (u_int32_t)uqmin((x), USHRT_MAX)
1602 #define SATU32(x) (u_int32_t)uqmin((x), UINT_MAX)
1603 
1604 	switch (version) {
1605 	case 0: {
1606 		struct pf_altq_v0 *exported_q =
1607 		    &((struct pfioc_altq_v0 *)pa)->altq;
1608 
1609 		COPY(ifname);
1610 
1611 		ASSIGN(scheduler);
1612 		ASSIGN(tbrsize);
1613 		exported_q->tbrsize = SATU16(q->tbrsize);
1614 		exported_q->ifbandwidth = SATU32(q->ifbandwidth);
1615 
1616 		COPY(qname);
1617 		COPY(parent);
1618 		ASSIGN(parent_qid);
1619 		exported_q->bandwidth = SATU32(q->bandwidth);
1620 		ASSIGN(priority);
1621 		ASSIGN(local_flags);
1622 
1623 		ASSIGN(qlimit);
1624 		ASSIGN(flags);
1625 
1626 		if (q->scheduler == ALTQT_HFSC) {
1627 #define ASSIGN_OPT(x) exported_q->pq_u.hfsc_opts.x = q->pq_u.hfsc_opts.x
1628 #define ASSIGN_OPT_SATU32(x) exported_q->pq_u.hfsc_opts.x = \
1629 			    SATU32(q->pq_u.hfsc_opts.x)
1630 
1631 			ASSIGN_OPT_SATU32(rtsc_m1);
1632 			ASSIGN_OPT(rtsc_d);
1633 			ASSIGN_OPT_SATU32(rtsc_m2);
1634 
1635 			ASSIGN_OPT_SATU32(lssc_m1);
1636 			ASSIGN_OPT(lssc_d);
1637 			ASSIGN_OPT_SATU32(lssc_m2);
1638 
1639 			ASSIGN_OPT_SATU32(ulsc_m1);
1640 			ASSIGN_OPT(ulsc_d);
1641 			ASSIGN_OPT_SATU32(ulsc_m2);
1642 
1643 			ASSIGN_OPT(flags);
1644 
1645 #undef ASSIGN_OPT
1646 #undef ASSIGN_OPT_SATU32
1647 		} else
1648 			COPY(pq_u);
1649 
1650 		ASSIGN(qid);
1651 		break;
1652 	}
1653 	case 1:	{
1654 		struct pf_altq_v1 *exported_q =
1655 		    &((struct pfioc_altq_v1 *)pa)->altq;
1656 
1657 		COPY(ifname);
1658 
1659 		ASSIGN(scheduler);
1660 		ASSIGN(tbrsize);
1661 		ASSIGN(ifbandwidth);
1662 
1663 		COPY(qname);
1664 		COPY(parent);
1665 		ASSIGN(parent_qid);
1666 		ASSIGN(bandwidth);
1667 		ASSIGN(priority);
1668 		ASSIGN(local_flags);
1669 
1670 		ASSIGN(qlimit);
1671 		ASSIGN(flags);
1672 		COPY(pq_u);
1673 
1674 		ASSIGN(qid);
1675 		break;
1676 	}
1677 	default:
1678 		panic("%s: unhandled struct pfioc_altq version", __func__);
1679 		break;
1680 	}
1681 
1682 #undef ASSIGN
1683 #undef COPY
1684 #undef SATU16
1685 #undef SATU32
1686 
1687 	return (0);
1688 }
1689 
1690 /*
1691  * Handle import to struct pf_kaltq of struct pf_altq from user binaries
1692  * that may be using any version of it.
1693  */
1694 static int
1695 pf_import_kaltq(struct pfioc_altq_v1 *pa, struct pf_altq *q, size_t ioc_size)
1696 {
1697 	u_int32_t version;
1698 
1699 	if (ioc_size == sizeof(struct pfioc_altq_v0))
1700 		version = 0;
1701 	else
1702 		version = pa->version;
1703 
1704 	if (version > PFIOC_ALTQ_VERSION)
1705 		return (EINVAL);
1706 
1707 #define ASSIGN(x) q->x = imported_q->x
1708 #define COPY(x) \
1709 	bcopy(&imported_q->x, &q->x, min(sizeof(imported_q->x), sizeof(q->x)))
1710 
1711 	switch (version) {
1712 	case 0: {
1713 		struct pf_altq_v0 *imported_q =
1714 		    &((struct pfioc_altq_v0 *)pa)->altq;
1715 
1716 		COPY(ifname);
1717 
1718 		ASSIGN(scheduler);
1719 		ASSIGN(tbrsize); /* 16-bit -> 32-bit */
1720 		ASSIGN(ifbandwidth); /* 32-bit -> 64-bit */
1721 
1722 		COPY(qname);
1723 		COPY(parent);
1724 		ASSIGN(parent_qid);
1725 		ASSIGN(bandwidth); /* 32-bit -> 64-bit */
1726 		ASSIGN(priority);
1727 		ASSIGN(local_flags);
1728 
1729 		ASSIGN(qlimit);
1730 		ASSIGN(flags);
1731 
1732 		if (imported_q->scheduler == ALTQT_HFSC) {
1733 #define ASSIGN_OPT(x) q->pq_u.hfsc_opts.x = imported_q->pq_u.hfsc_opts.x
1734 
1735 			/*
1736 			 * The m1 and m2 parameters are being copied from
1737 			 * 32-bit to 64-bit.
1738 			 */
1739 			ASSIGN_OPT(rtsc_m1);
1740 			ASSIGN_OPT(rtsc_d);
1741 			ASSIGN_OPT(rtsc_m2);
1742 
1743 			ASSIGN_OPT(lssc_m1);
1744 			ASSIGN_OPT(lssc_d);
1745 			ASSIGN_OPT(lssc_m2);
1746 
1747 			ASSIGN_OPT(ulsc_m1);
1748 			ASSIGN_OPT(ulsc_d);
1749 			ASSIGN_OPT(ulsc_m2);
1750 
1751 			ASSIGN_OPT(flags);
1752 
1753 #undef ASSIGN_OPT
1754 		} else
1755 			COPY(pq_u);
1756 
1757 		ASSIGN(qid);
1758 		break;
1759 	}
1760 	case 1: {
1761 		struct pf_altq_v1 *imported_q =
1762 		    &((struct pfioc_altq_v1 *)pa)->altq;
1763 
1764 		COPY(ifname);
1765 
1766 		ASSIGN(scheduler);
1767 		ASSIGN(tbrsize);
1768 		ASSIGN(ifbandwidth);
1769 
1770 		COPY(qname);
1771 		COPY(parent);
1772 		ASSIGN(parent_qid);
1773 		ASSIGN(bandwidth);
1774 		ASSIGN(priority);
1775 		ASSIGN(local_flags);
1776 
1777 		ASSIGN(qlimit);
1778 		ASSIGN(flags);
1779 		COPY(pq_u);
1780 
1781 		ASSIGN(qid);
1782 		break;
1783 	}
1784 	default:
1785 		panic("%s: unhandled struct pfioc_altq version", __func__);
1786 		break;
1787 	}
1788 
1789 #undef ASSIGN
1790 #undef COPY
1791 
1792 	return (0);
1793 }
1794 
1795 static struct pf_altq *
1796 pf_altq_get_nth_active(u_int32_t n)
1797 {
1798 	struct pf_altq		*altq;
1799 	u_int32_t		 nr;
1800 
1801 	nr = 0;
1802 	TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
1803 		if (nr == n)
1804 			return (altq);
1805 		nr++;
1806 	}
1807 
1808 	TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
1809 		if (nr == n)
1810 			return (altq);
1811 		nr++;
1812 	}
1813 
1814 	return (NULL);
1815 }
1816 #endif /* ALTQ */
1817 
1818 struct pf_krule *
1819 pf_krule_alloc(void)
1820 {
1821 	struct pf_krule *rule;
1822 
1823 	rule = malloc(sizeof(struct pf_krule), M_PFRULE, M_WAITOK | M_ZERO);
1824 	mtx_init(&rule->rpool.mtx, "pf_krule_pool", NULL, MTX_DEF);
1825 	rule->timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
1826 	    M_WAITOK | M_ZERO);
1827 	return (rule);
1828 }
1829 
1830 void
1831 pf_krule_free(struct pf_krule *rule)
1832 {
1833 #ifdef PF_WANT_32_TO_64_COUNTER
1834 	bool wowned;
1835 #endif
1836 
1837 	if (rule == NULL)
1838 		return;
1839 
1840 #ifdef PF_WANT_32_TO_64_COUNTER
1841 	if (rule->allrulelinked) {
1842 		wowned = PF_RULES_WOWNED();
1843 		if (!wowned)
1844 			PF_RULES_WLOCK();
1845 		LIST_REMOVE(rule, allrulelist);
1846 		V_pf_allrulecount--;
1847 		if (!wowned)
1848 			PF_RULES_WUNLOCK();
1849 	}
1850 #endif
1851 
1852 	pf_counter_u64_deinit(&rule->evaluations);
1853 	for (int i = 0; i < 2; i++) {
1854 		pf_counter_u64_deinit(&rule->packets[i]);
1855 		pf_counter_u64_deinit(&rule->bytes[i]);
1856 	}
1857 	counter_u64_free(rule->states_cur);
1858 	counter_u64_free(rule->states_tot);
1859 	counter_u64_free(rule->src_nodes);
1860 	uma_zfree_pcpu(pf_timestamp_pcpu_zone, rule->timestamp);
1861 
1862 	mtx_destroy(&rule->rpool.mtx);
1863 	free(rule, M_PFRULE);
1864 }
1865 
1866 void
1867 pf_krule_clear_counters(struct pf_krule *rule)
1868 {
1869 	pf_counter_u64_zero(&rule->evaluations);
1870 	for (int i = 0; i < 2; i++) {
1871 		pf_counter_u64_zero(&rule->packets[i]);
1872 		pf_counter_u64_zero(&rule->bytes[i]);
1873 	}
1874 	counter_u64_zero(rule->states_tot);
1875 }
1876 
1877 static void
1878 pf_kpooladdr_to_pooladdr(const struct pf_kpooladdr *kpool,
1879     struct pf_pooladdr *pool)
1880 {
1881 
1882 	bzero(pool, sizeof(*pool));
1883 	bcopy(&kpool->addr, &pool->addr, sizeof(pool->addr));
1884 	strlcpy(pool->ifname, kpool->ifname, sizeof(pool->ifname));
1885 }
1886 
1887 static int
1888 pf_pooladdr_to_kpooladdr(const struct pf_pooladdr *pool,
1889     struct pf_kpooladdr *kpool)
1890 {
1891 	int ret;
1892 
1893 	bzero(kpool, sizeof(*kpool));
1894 	bcopy(&pool->addr, &kpool->addr, sizeof(kpool->addr));
1895 	ret = pf_user_strcpy(kpool->ifname, pool->ifname,
1896 	    sizeof(kpool->ifname));
1897 	return (ret);
1898 }
1899 
1900 static void
1901 pf_pool_to_kpool(const struct pf_pool *pool, struct pf_kpool *kpool)
1902 {
1903 	_Static_assert(sizeof(pool->key) == sizeof(kpool->key), "");
1904 	_Static_assert(sizeof(pool->counter) == sizeof(kpool->counter), "");
1905 
1906 	bcopy(&pool->key, &kpool->key, sizeof(kpool->key));
1907 	bcopy(&pool->counter, &kpool->counter, sizeof(kpool->counter));
1908 
1909 	kpool->tblidx = pool->tblidx;
1910 	kpool->proxy_port[0] = pool->proxy_port[0];
1911 	kpool->proxy_port[1] = pool->proxy_port[1];
1912 	kpool->opts = pool->opts;
1913 }
1914 
1915 static int
1916 pf_rule_to_krule(const struct pf_rule *rule, struct pf_krule *krule)
1917 {
1918 	int ret;
1919 
1920 #ifndef INET
1921 	if (rule->af == AF_INET) {
1922 		return (EAFNOSUPPORT);
1923 	}
1924 #endif /* INET */
1925 #ifndef INET6
1926 	if (rule->af == AF_INET6) {
1927 		return (EAFNOSUPPORT);
1928 	}
1929 #endif /* INET6 */
1930 
1931 	ret = pf_check_rule_addr(&rule->src);
1932 	if (ret != 0)
1933 		return (ret);
1934 	ret = pf_check_rule_addr(&rule->dst);
1935 	if (ret != 0)
1936 		return (ret);
1937 
1938 	bcopy(&rule->src, &krule->src, sizeof(rule->src));
1939 	bcopy(&rule->dst, &krule->dst, sizeof(rule->dst));
1940 
1941 	ret = pf_user_strcpy(krule->label[0], rule->label, sizeof(rule->label));
1942 	if (ret != 0)
1943 		return (ret);
1944 	ret = pf_user_strcpy(krule->ifname, rule->ifname, sizeof(rule->ifname));
1945 	if (ret != 0)
1946 		return (ret);
1947 	ret = pf_user_strcpy(krule->qname, rule->qname, sizeof(rule->qname));
1948 	if (ret != 0)
1949 		return (ret);
1950 	ret = pf_user_strcpy(krule->pqname, rule->pqname, sizeof(rule->pqname));
1951 	if (ret != 0)
1952 		return (ret);
1953 	ret = pf_user_strcpy(krule->tagname, rule->tagname,
1954 	    sizeof(rule->tagname));
1955 	if (ret != 0)
1956 		return (ret);
1957 	ret = pf_user_strcpy(krule->match_tagname, rule->match_tagname,
1958 	    sizeof(rule->match_tagname));
1959 	if (ret != 0)
1960 		return (ret);
1961 	ret = pf_user_strcpy(krule->overload_tblname, rule->overload_tblname,
1962 	    sizeof(rule->overload_tblname));
1963 	if (ret != 0)
1964 		return (ret);
1965 
1966 	pf_pool_to_kpool(&rule->rpool, &krule->rpool);
1967 
1968 	/* Don't allow userspace to set evaluations, packets or bytes. */
1969 	/* kif, anchor, overload_tbl are not copied over. */
1970 
1971 	krule->os_fingerprint = rule->os_fingerprint;
1972 
1973 	krule->rtableid = rule->rtableid;
1974 	/* pf_rule->timeout is smaller than pf_krule->timeout */
1975 	bcopy(rule->timeout, krule->timeout, sizeof(rule->timeout));
1976 	krule->max_states = rule->max_states;
1977 	krule->max_src_nodes = rule->max_src_nodes;
1978 	krule->max_src_states = rule->max_src_states;
1979 	krule->max_src_conn = rule->max_src_conn;
1980 	krule->max_src_conn_rate.limit = rule->max_src_conn_rate.limit;
1981 	krule->max_src_conn_rate.seconds = rule->max_src_conn_rate.seconds;
1982 	krule->qid = rule->qid;
1983 	krule->pqid = rule->pqid;
1984 	krule->nr = rule->nr;
1985 	krule->prob = rule->prob;
1986 	krule->cuid = rule->cuid;
1987 	krule->cpid = rule->cpid;
1988 
1989 	krule->return_icmp = rule->return_icmp;
1990 	krule->return_icmp6 = rule->return_icmp6;
1991 	krule->max_mss = rule->max_mss;
1992 	krule->tag = rule->tag;
1993 	krule->match_tag = rule->match_tag;
1994 	krule->scrub_flags = rule->scrub_flags;
1995 
1996 	bcopy(&rule->uid, &krule->uid, sizeof(krule->uid));
1997 	bcopy(&rule->gid, &krule->gid, sizeof(krule->gid));
1998 
1999 	krule->rule_flag = rule->rule_flag;
2000 	krule->action = rule->action;
2001 	krule->direction = rule->direction;
2002 	krule->log = rule->log;
2003 	krule->logif = rule->logif;
2004 	krule->quick = rule->quick;
2005 	krule->ifnot = rule->ifnot;
2006 	krule->match_tag_not = rule->match_tag_not;
2007 	krule->natpass = rule->natpass;
2008 
2009 	krule->keep_state = rule->keep_state;
2010 	krule->af = rule->af;
2011 	krule->proto = rule->proto;
2012 	krule->type = rule->type;
2013 	krule->code = rule->code;
2014 	krule->flags = rule->flags;
2015 	krule->flagset = rule->flagset;
2016 	krule->min_ttl = rule->min_ttl;
2017 	krule->allow_opts = rule->allow_opts;
2018 	krule->rt = rule->rt;
2019 	krule->return_ttl = rule->return_ttl;
2020 	krule->tos = rule->tos;
2021 	krule->set_tos = rule->set_tos;
2022 
2023 	krule->flush = rule->flush;
2024 	krule->prio = rule->prio;
2025 	krule->set_prio[0] = rule->set_prio[0];
2026 	krule->set_prio[1] = rule->set_prio[1];
2027 
2028 	bcopy(&rule->divert, &krule->divert, sizeof(krule->divert));
2029 
2030 	return (0);
2031 }
2032 
2033 int
2034 pf_ioctl_getrules(struct pfioc_rule *pr)
2035 {
2036 	struct pf_kruleset	*ruleset;
2037 	struct pf_krule		*tail;
2038 	int			 rs_num;
2039 
2040 	PF_RULES_WLOCK();
2041 	ruleset = pf_find_kruleset(pr->anchor);
2042 	if (ruleset == NULL) {
2043 		PF_RULES_WUNLOCK();
2044 		return (EINVAL);
2045 	}
2046 	rs_num = pf_get_ruleset_number(pr->rule.action);
2047 	if (rs_num >= PF_RULESET_MAX) {
2048 		PF_RULES_WUNLOCK();
2049 		return (EINVAL);
2050 	}
2051 	tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
2052 	    pf_krulequeue);
2053 	if (tail)
2054 		pr->nr = tail->nr + 1;
2055 	else
2056 		pr->nr = 0;
2057 	pr->ticket = ruleset->rules[rs_num].active.ticket;
2058 	PF_RULES_WUNLOCK();
2059 
2060 	return (0);
2061 }
2062 
2063 int
2064 pf_ioctl_addrule(struct pf_krule *rule, uint32_t ticket,
2065     uint32_t pool_ticket, const char *anchor, const char *anchor_call,
2066     uid_t uid, pid_t pid)
2067 {
2068 	struct pf_kruleset	*ruleset;
2069 	struct pf_krule		*tail;
2070 	struct pf_kpooladdr	*pa;
2071 	struct pfi_kkif		*kif = NULL;
2072 	int			 rs_num;
2073 	int			 error = 0;
2074 
2075 	if ((rule->return_icmp >> 8) > ICMP_MAXTYPE) {
2076 		error = EINVAL;
2077 		goto errout_unlocked;
2078 	}
2079 
2080 #define	ERROUT(x)	ERROUT_FUNCTION(errout, x)
2081 
2082 	if (rule->ifname[0])
2083 		kif = pf_kkif_create(M_WAITOK);
2084 	pf_counter_u64_init(&rule->evaluations, M_WAITOK);
2085 	for (int i = 0; i < 2; i++) {
2086 		pf_counter_u64_init(&rule->packets[i], M_WAITOK);
2087 		pf_counter_u64_init(&rule->bytes[i], M_WAITOK);
2088 	}
2089 	rule->states_cur = counter_u64_alloc(M_WAITOK);
2090 	rule->states_tot = counter_u64_alloc(M_WAITOK);
2091 	rule->src_nodes = counter_u64_alloc(M_WAITOK);
2092 	rule->cuid = uid;
2093 	rule->cpid = pid;
2094 	TAILQ_INIT(&rule->rpool.list);
2095 
2096 	PF_CONFIG_LOCK();
2097 	PF_RULES_WLOCK();
2098 #ifdef PF_WANT_32_TO_64_COUNTER
2099 	LIST_INSERT_HEAD(&V_pf_allrulelist, rule, allrulelist);
2100 	MPASS(!rule->allrulelinked);
2101 	rule->allrulelinked = true;
2102 	V_pf_allrulecount++;
2103 #endif
2104 	ruleset = pf_find_kruleset(anchor);
2105 	if (ruleset == NULL)
2106 		ERROUT(EINVAL);
2107 	rs_num = pf_get_ruleset_number(rule->action);
2108 	if (rs_num >= PF_RULESET_MAX)
2109 		ERROUT(EINVAL);
2110 	if (ticket != ruleset->rules[rs_num].inactive.ticket) {
2111 		DPFPRINTF(PF_DEBUG_MISC,
2112 		    ("ticket: %d != [%d]%d\n", ticket, rs_num,
2113 		    ruleset->rules[rs_num].inactive.ticket));
2114 		ERROUT(EBUSY);
2115 	}
2116 	if (pool_ticket != V_ticket_pabuf) {
2117 		DPFPRINTF(PF_DEBUG_MISC,
2118 		    ("pool_ticket: %d != %d\n", pool_ticket,
2119 		    V_ticket_pabuf));
2120 		ERROUT(EBUSY);
2121 	}
2122 	/*
2123 	 * XXXMJG hack: there is no mechanism to ensure they started the
2124 	 * transaction. Ticket checked above may happen to match by accident,
2125 	 * even if nobody called DIOCXBEGIN, let alone this process.
2126 	 * Partially work around it by checking if the RB tree got allocated,
2127 	 * see pf_begin_rules.
2128 	 */
2129 	if (ruleset->rules[rs_num].inactive.tree == NULL) {
2130 		ERROUT(EINVAL);
2131 	}
2132 
2133 	tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
2134 	    pf_krulequeue);
2135 	if (tail)
2136 		rule->nr = tail->nr + 1;
2137 	else
2138 		rule->nr = 0;
2139 	if (rule->ifname[0]) {
2140 		rule->kif = pfi_kkif_attach(kif, rule->ifname);
2141 		kif = NULL;
2142 		pfi_kkif_ref(rule->kif);
2143 	} else
2144 		rule->kif = NULL;
2145 
2146 	if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs)
2147 		error = EBUSY;
2148 
2149 #ifdef ALTQ
2150 	/* set queue IDs */
2151 	if (rule->qname[0] != 0) {
2152 		if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
2153 			error = EBUSY;
2154 		else if (rule->pqname[0] != 0) {
2155 			if ((rule->pqid =
2156 			    pf_qname2qid(rule->pqname)) == 0)
2157 				error = EBUSY;
2158 		} else
2159 			rule->pqid = rule->qid;
2160 	}
2161 #endif
2162 	if (rule->tagname[0])
2163 		if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
2164 			error = EBUSY;
2165 	if (rule->match_tagname[0])
2166 		if ((rule->match_tag =
2167 		    pf_tagname2tag(rule->match_tagname)) == 0)
2168 			error = EBUSY;
2169 	if (rule->rt && !rule->direction)
2170 		error = EINVAL;
2171 	if (!rule->log)
2172 		rule->logif = 0;
2173 	if (rule->logif >= PFLOGIFS_MAX)
2174 		error = EINVAL;
2175 	if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
2176 		error = ENOMEM;
2177 	if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
2178 		error = ENOMEM;
2179 	if (pf_kanchor_setup(rule, ruleset, anchor_call))
2180 		error = EINVAL;
2181 	if (rule->scrub_flags & PFSTATE_SETPRIO &&
2182 	    (rule->set_prio[0] > PF_PRIO_MAX ||
2183 	    rule->set_prio[1] > PF_PRIO_MAX))
2184 		error = EINVAL;
2185 	TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
2186 		if (pa->addr.type == PF_ADDR_TABLE) {
2187 			pa->addr.p.tbl = pfr_attach_table(ruleset,
2188 			    pa->addr.v.tblname);
2189 			if (pa->addr.p.tbl == NULL)
2190 				error = ENOMEM;
2191 		}
2192 
2193 	rule->overload_tbl = NULL;
2194 	if (rule->overload_tblname[0]) {
2195 		if ((rule->overload_tbl = pfr_attach_table(ruleset,
2196 		    rule->overload_tblname)) == NULL)
2197 			error = EINVAL;
2198 		else
2199 			rule->overload_tbl->pfrkt_flags |=
2200 			    PFR_TFLAG_ACTIVE;
2201 	}
2202 
2203 	pf_mv_kpool(&V_pf_pabuf, &rule->rpool.list);
2204 	if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
2205 	    (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
2206 	    (rule->rt > PF_NOPFROUTE)) &&
2207 	    (TAILQ_FIRST(&rule->rpool.list) == NULL))
2208 		error = EINVAL;
2209 
2210 	if (rule->action == PF_PASS && rule->rpool.opts & PF_POOL_STICKYADDR &&
2211 	    !rule->keep_state) {
2212 		error = EINVAL;
2213 	}
2214 
2215 	if (error) {
2216 		pf_free_rule(rule);
2217 		rule = NULL;
2218 		ERROUT(error);
2219 	}
2220 
2221 	rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
2222 	TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
2223 	    rule, entries);
2224 	ruleset->rules[rs_num].inactive.rcount++;
2225 
2226 	PF_RULES_WUNLOCK();
2227 	pf_hash_rule(rule);
2228 	if (RB_INSERT(pf_krule_global, ruleset->rules[rs_num].inactive.tree, rule) != NULL) {
2229 		PF_RULES_WLOCK();
2230 		TAILQ_REMOVE(ruleset->rules[rs_num].inactive.ptr, rule, entries);
2231 		ruleset->rules[rs_num].inactive.rcount--;
2232 		pf_free_rule(rule);
2233 		rule = NULL;
2234 		ERROUT(EEXIST);
2235 	}
2236 	PF_CONFIG_UNLOCK();
2237 
2238 	return (0);
2239 
2240 #undef ERROUT
2241 errout:
2242 	PF_RULES_WUNLOCK();
2243 	PF_CONFIG_UNLOCK();
2244 errout_unlocked:
2245 	pf_kkif_free(kif);
2246 	pf_krule_free(rule);
2247 	return (error);
2248 }
2249 
2250 static bool
2251 pf_label_match(const struct pf_krule *rule, const char *label)
2252 {
2253 	int i = 0;
2254 
2255 	while (*rule->label[i]) {
2256 		if (strcmp(rule->label[i], label) == 0)
2257 			return (true);
2258 		i++;
2259 	}
2260 
2261 	return (false);
2262 }
2263 
2264 static unsigned int
2265 pf_kill_matching_state(struct pf_state_key_cmp *key, int dir)
2266 {
2267 	struct pf_kstate *s;
2268 	int more = 0;
2269 
2270 	s = pf_find_state_all(key, dir, &more);
2271 	if (s == NULL)
2272 		return (0);
2273 
2274 	if (more) {
2275 		PF_STATE_UNLOCK(s);
2276 		return (0);
2277 	}
2278 
2279 	pf_unlink_state(s);
2280 	return (1);
2281 }
2282 
2283 static int
2284 pf_killstates_row(struct pf_kstate_kill *psk, struct pf_idhash *ih)
2285 {
2286 	struct pf_kstate	*s;
2287 	struct pf_state_key	*sk;
2288 	struct pf_addr		*srcaddr, *dstaddr;
2289 	struct pf_state_key_cmp	 match_key;
2290 	int			 idx, killed = 0;
2291 	unsigned int		 dir;
2292 	u_int16_t		 srcport, dstport;
2293 	struct pfi_kkif		*kif;
2294 
2295 relock_DIOCKILLSTATES:
2296 	PF_HASHROW_LOCK(ih);
2297 	LIST_FOREACH(s, &ih->states, entry) {
2298 		/* For floating states look at the original kif. */
2299 		kif = s->kif == V_pfi_all ? s->orig_kif : s->kif;
2300 
2301 		sk = s->key[psk->psk_nat ? PF_SK_STACK : PF_SK_WIRE];
2302 		if (s->direction == PF_OUT) {
2303 			srcaddr = &sk->addr[1];
2304 			dstaddr = &sk->addr[0];
2305 			srcport = sk->port[1];
2306 			dstport = sk->port[0];
2307 		} else {
2308 			srcaddr = &sk->addr[0];
2309 			dstaddr = &sk->addr[1];
2310 			srcport = sk->port[0];
2311 			dstport = sk->port[1];
2312 		}
2313 
2314 		if (psk->psk_af && sk->af != psk->psk_af)
2315 			continue;
2316 
2317 		if (psk->psk_proto && psk->psk_proto != sk->proto)
2318 			continue;
2319 
2320 		if (! PF_MATCHA(psk->psk_src.neg, &psk->psk_src.addr.v.a.addr,
2321 		    &psk->psk_src.addr.v.a.mask, srcaddr, sk->af))
2322 			continue;
2323 
2324 		if (! PF_MATCHA(psk->psk_dst.neg, &psk->psk_dst.addr.v.a.addr,
2325 		    &psk->psk_dst.addr.v.a.mask, dstaddr, sk->af))
2326 			continue;
2327 
2328 		if (!  PF_MATCHA(psk->psk_rt_addr.neg,
2329 		    &psk->psk_rt_addr.addr.v.a.addr,
2330 		    &psk->psk_rt_addr.addr.v.a.mask,
2331 		    &s->rt_addr, sk->af))
2332 			continue;
2333 
2334 		if (psk->psk_src.port_op != 0 &&
2335 		    ! pf_match_port(psk->psk_src.port_op,
2336 		    psk->psk_src.port[0], psk->psk_src.port[1], srcport))
2337 			continue;
2338 
2339 		if (psk->psk_dst.port_op != 0 &&
2340 		    ! pf_match_port(psk->psk_dst.port_op,
2341 		    psk->psk_dst.port[0], psk->psk_dst.port[1], dstport))
2342 			continue;
2343 
2344 		if (psk->psk_label[0] &&
2345 		    ! pf_label_match(s->rule.ptr, psk->psk_label))
2346 			continue;
2347 
2348 		if (psk->psk_ifname[0] && strcmp(psk->psk_ifname,
2349 		    kif->pfik_name))
2350 			continue;
2351 
2352 		if (psk->psk_kill_match) {
2353 			/* Create the key to find matching states, with lock
2354 			 * held. */
2355 
2356 			bzero(&match_key, sizeof(match_key));
2357 
2358 			if (s->direction == PF_OUT) {
2359 				dir = PF_IN;
2360 				idx = psk->psk_nat ? PF_SK_WIRE : PF_SK_STACK;
2361 			} else {
2362 				dir = PF_OUT;
2363 				idx = psk->psk_nat ? PF_SK_STACK : PF_SK_WIRE;
2364 			}
2365 
2366 			match_key.af = s->key[idx]->af;
2367 			match_key.proto = s->key[idx]->proto;
2368 			PF_ACPY(&match_key.addr[0],
2369 			    &s->key[idx]->addr[1], match_key.af);
2370 			match_key.port[0] = s->key[idx]->port[1];
2371 			PF_ACPY(&match_key.addr[1],
2372 			    &s->key[idx]->addr[0], match_key.af);
2373 			match_key.port[1] = s->key[idx]->port[0];
2374 		}
2375 
2376 		pf_unlink_state(s);
2377 		killed++;
2378 
2379 		if (psk->psk_kill_match)
2380 			killed += pf_kill_matching_state(&match_key, dir);
2381 
2382 		goto relock_DIOCKILLSTATES;
2383 	}
2384 	PF_HASHROW_UNLOCK(ih);
2385 
2386 	return (killed);
2387 }
2388 
2389 int
2390 pf_start(void)
2391 {
2392 	int error = 0;
2393 
2394 	sx_xlock(&V_pf_ioctl_lock);
2395 	if (V_pf_status.running)
2396 		error = EEXIST;
2397 	else {
2398 		hook_pf();
2399 		if (! TAILQ_EMPTY(V_pf_keth->active.rules))
2400 			hook_pf_eth();
2401 		V_pf_status.running = 1;
2402 		V_pf_status.since = time_second;
2403 		new_unrhdr64(&V_pf_stateid, time_second);
2404 
2405 		DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
2406 	}
2407 	sx_xunlock(&V_pf_ioctl_lock);
2408 
2409 	return (error);
2410 }
2411 
2412 int
2413 pf_stop(void)
2414 {
2415 	int error = 0;
2416 
2417 	sx_xlock(&V_pf_ioctl_lock);
2418 	if (!V_pf_status.running)
2419 		error = ENOENT;
2420 	else {
2421 		V_pf_status.running = 0;
2422 		dehook_pf();
2423 		dehook_pf_eth();
2424 		V_pf_status.since = time_second;
2425 		DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
2426 	}
2427 	sx_xunlock(&V_pf_ioctl_lock);
2428 
2429 	return (error);
2430 }
2431 
2432 void
2433 pf_ioctl_clear_status(void)
2434 {
2435 	PF_RULES_WLOCK();
2436 	for (int i = 0; i < PFRES_MAX; i++)
2437 		counter_u64_zero(V_pf_status.counters[i]);
2438 	for (int i = 0; i < FCNT_MAX; i++)
2439 		pf_counter_u64_zero(&V_pf_status.fcounters[i]);
2440 	for (int i = 0; i < SCNT_MAX; i++)
2441 		counter_u64_zero(V_pf_status.scounters[i]);
2442 	for (int i = 0; i < KLCNT_MAX; i++)
2443 		counter_u64_zero(V_pf_status.lcounters[i]);
2444 	V_pf_status.since = time_second;
2445 	if (*V_pf_status.ifname)
2446 		pfi_update_status(V_pf_status.ifname, NULL);
2447 	PF_RULES_WUNLOCK();
2448 }
2449 
2450 int
2451 pf_ioctl_set_timeout(int timeout, int seconds, int *prev_seconds)
2452 {
2453 	uint32_t old;
2454 
2455 	if (timeout < 0 || timeout >= PFTM_MAX ||
2456 	    seconds < 0)
2457 		return (EINVAL);
2458 
2459 	PF_RULES_WLOCK();
2460 	old = V_pf_default_rule.timeout[timeout];
2461 	if (timeout == PFTM_INTERVAL && seconds == 0)
2462 		seconds = 1;
2463 	V_pf_default_rule.timeout[timeout] = seconds;
2464 	if (timeout == PFTM_INTERVAL && seconds < old)
2465 		wakeup(pf_purge_thread);
2466 
2467 	if (prev_seconds != NULL)
2468 		*prev_seconds = old;
2469 
2470 	PF_RULES_WUNLOCK();
2471 
2472 	return (0);
2473 }
2474 
2475 int
2476 pf_ioctl_get_timeout(int timeout, int *seconds)
2477 {
2478 	PF_RULES_RLOCK_TRACKER;
2479 
2480 	if (timeout < 0 || timeout >= PFTM_MAX)
2481 		return (EINVAL);
2482 
2483 	PF_RULES_RLOCK();
2484 	*seconds = V_pf_default_rule.timeout[timeout];
2485 	PF_RULES_RUNLOCK();
2486 
2487 	return (0);
2488 }
2489 
2490 int
2491 pf_ioctl_set_limit(int index, unsigned int limit, unsigned int *old_limit)
2492 {
2493 
2494 	PF_RULES_WLOCK();
2495 	if (index < 0 || index >= PF_LIMIT_MAX ||
2496 	    V_pf_limits[index].zone == NULL) {
2497 		PF_RULES_WUNLOCK();
2498 		return (EINVAL);
2499 	}
2500 	uma_zone_set_max(V_pf_limits[index].zone, limit);
2501 	if (old_limit != NULL)
2502 		*old_limit = V_pf_limits[index].limit;
2503 	V_pf_limits[index].limit = limit;
2504 	PF_RULES_WUNLOCK();
2505 
2506 	return (0);
2507 }
2508 
2509 int
2510 pf_ioctl_get_limit(int index, unsigned int *limit)
2511 {
2512 	PF_RULES_RLOCK_TRACKER;
2513 
2514 	if (index < 0 || index >= PF_LIMIT_MAX)
2515 		return (EINVAL);
2516 
2517 	PF_RULES_RLOCK();
2518 	*limit = V_pf_limits[index].limit;
2519 	PF_RULES_RUNLOCK();
2520 
2521 	return (0);
2522 }
2523 
2524 int
2525 pf_ioctl_begin_addrs(uint32_t *ticket)
2526 {
2527 	PF_RULES_WLOCK();
2528 	pf_empty_kpool(&V_pf_pabuf);
2529 	*ticket = ++V_ticket_pabuf;
2530 	PF_RULES_WUNLOCK();
2531 
2532 	return (0);
2533 }
2534 
2535 int
2536 pf_ioctl_add_addr(struct pfioc_pooladdr *pp)
2537 {
2538 	struct pf_kpooladdr	*pa = NULL;
2539 	struct pfi_kkif		*kif = NULL;
2540 	int error;
2541 
2542 #ifndef INET
2543 	if (pp->af == AF_INET)
2544 		return (EAFNOSUPPORT);
2545 #endif /* INET */
2546 #ifndef INET6
2547 	if (pp->af == AF_INET6)
2548 		return (EAFNOSUPPORT);
2549 #endif /* INET6 */
2550 
2551 	if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2552 	    pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2553 	    pp->addr.addr.type != PF_ADDR_TABLE)
2554 		return (EINVAL);
2555 
2556 	if (pp->addr.addr.p.dyn != NULL)
2557 		return (EINVAL);
2558 
2559 	pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
2560 	error = pf_pooladdr_to_kpooladdr(&pp->addr, pa);
2561 	if (error != 0)
2562 		goto out;
2563 	if (pa->ifname[0])
2564 		kif = pf_kkif_create(M_WAITOK);
2565 	PF_RULES_WLOCK();
2566 	if (pp->ticket != V_ticket_pabuf) {
2567 		PF_RULES_WUNLOCK();
2568 		if (pa->ifname[0])
2569 			pf_kkif_free(kif);
2570 		error = EBUSY;
2571 		goto out;
2572 	}
2573 	if (pa->ifname[0]) {
2574 		pa->kif = pfi_kkif_attach(kif, pa->ifname);
2575 		kif = NULL;
2576 		pfi_kkif_ref(pa->kif);
2577 	} else
2578 		pa->kif = NULL;
2579 	if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
2580 	    pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
2581 		if (pa->ifname[0])
2582 			pfi_kkif_unref(pa->kif);
2583 		PF_RULES_WUNLOCK();
2584 		goto out;
2585 	}
2586 	TAILQ_INSERT_TAIL(&V_pf_pabuf, pa, entries);
2587 	PF_RULES_WUNLOCK();
2588 
2589 	return (0);
2590 
2591 out:
2592 	free(pa, M_PFRULE);
2593 	return (error);
2594 }
2595 
2596 int
2597 pf_ioctl_get_addrs(struct pfioc_pooladdr *pp)
2598 {
2599 	struct pf_kpool		*pool;
2600 	struct pf_kpooladdr	*pa;
2601 
2602 	PF_RULES_RLOCK_TRACKER;
2603 
2604 	pp->anchor[sizeof(pp->anchor) - 1] = 0;
2605 	pp->nr = 0;
2606 
2607 	PF_RULES_RLOCK();
2608 	pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
2609 	    pp->r_num, 0, 1, 0);
2610 	if (pool == NULL) {
2611 		PF_RULES_RUNLOCK();
2612 		return (EBUSY);
2613 	}
2614 	TAILQ_FOREACH(pa, &pool->list, entries)
2615 		pp->nr++;
2616 	PF_RULES_RUNLOCK();
2617 
2618 	return (0);
2619 }
2620 
2621 int
2622 pf_ioctl_get_addr(struct pfioc_pooladdr *pp)
2623 {
2624 	struct pf_kpool		*pool;
2625 	struct pf_kpooladdr	*pa;
2626 	u_int32_t		 nr = 0;
2627 
2628 	PF_RULES_RLOCK_TRACKER;
2629 
2630 	pp->anchor[sizeof(pp->anchor) - 1] = 0;
2631 
2632 	PF_RULES_RLOCK();
2633 	pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
2634 	    pp->r_num, 0, 1, 1);
2635 	if (pool == NULL) {
2636 		PF_RULES_RUNLOCK();
2637 		return (EBUSY);
2638 	}
2639 	pa = TAILQ_FIRST(&pool->list);
2640 	while ((pa != NULL) && (nr < pp->nr)) {
2641 		pa = TAILQ_NEXT(pa, entries);
2642 		nr++;
2643 	}
2644 	if (pa == NULL) {
2645 		PF_RULES_RUNLOCK();
2646 		return (EBUSY);
2647 	}
2648 	pf_kpooladdr_to_pooladdr(pa, &pp->addr);
2649 	pf_addr_copyout(&pp->addr.addr);
2650 	PF_RULES_RUNLOCK();
2651 
2652 	return (0);
2653 }
2654 
2655 static int
2656 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
2657 {
2658 	int			 error = 0;
2659 	PF_RULES_RLOCK_TRACKER;
2660 
2661 #define	ERROUT_IOCTL(target, x)					\
2662     do {								\
2663 	    error = (x);						\
2664 	    SDT_PROBE3(pf, ioctl, ioctl, error, cmd, error, __LINE__);	\
2665 	    goto target;						\
2666     } while (0)
2667 
2668 
2669 	/* XXX keep in sync with switch() below */
2670 	if (securelevel_gt(td->td_ucred, 2))
2671 		switch (cmd) {
2672 		case DIOCGETRULES:
2673 		case DIOCGETRULENV:
2674 		case DIOCGETADDRS:
2675 		case DIOCGETADDR:
2676 		case DIOCGETSTATE:
2677 		case DIOCGETSTATENV:
2678 		case DIOCSETSTATUSIF:
2679 		case DIOCGETSTATUSNV:
2680 		case DIOCCLRSTATUS:
2681 		case DIOCNATLOOK:
2682 		case DIOCSETDEBUG:
2683 #ifdef COMPAT_FREEBSD14
2684 		case DIOCGETSTATES:
2685 		case DIOCGETSTATESV2:
2686 #endif
2687 		case DIOCGETTIMEOUT:
2688 		case DIOCCLRRULECTRS:
2689 		case DIOCGETLIMIT:
2690 		case DIOCGETALTQSV0:
2691 		case DIOCGETALTQSV1:
2692 		case DIOCGETALTQV0:
2693 		case DIOCGETALTQV1:
2694 		case DIOCGETQSTATSV0:
2695 		case DIOCGETQSTATSV1:
2696 		case DIOCGETRULESETS:
2697 		case DIOCGETRULESET:
2698 		case DIOCRGETTABLES:
2699 		case DIOCRGETTSTATS:
2700 		case DIOCRCLRTSTATS:
2701 		case DIOCRCLRADDRS:
2702 		case DIOCRADDADDRS:
2703 		case DIOCRDELADDRS:
2704 		case DIOCRSETADDRS:
2705 		case DIOCRGETADDRS:
2706 		case DIOCRGETASTATS:
2707 		case DIOCRCLRASTATS:
2708 		case DIOCRTSTADDRS:
2709 		case DIOCOSFPGET:
2710 		case DIOCGETSRCNODES:
2711 		case DIOCCLRSRCNODES:
2712 		case DIOCGETSYNCOOKIES:
2713 		case DIOCIGETIFACES:
2714 		case DIOCGIFSPEEDV0:
2715 		case DIOCGIFSPEEDV1:
2716 		case DIOCSETIFFLAG:
2717 		case DIOCCLRIFFLAG:
2718 		case DIOCGETETHRULES:
2719 		case DIOCGETETHRULE:
2720 		case DIOCGETETHRULESETS:
2721 		case DIOCGETETHRULESET:
2722 			break;
2723 		case DIOCRCLRTABLES:
2724 		case DIOCRADDTABLES:
2725 		case DIOCRDELTABLES:
2726 		case DIOCRSETTFLAGS:
2727 			if (((struct pfioc_table *)addr)->pfrio_flags &
2728 			    PFR_FLAG_DUMMY)
2729 				break; /* dummy operation ok */
2730 			return (EPERM);
2731 		default:
2732 			return (EPERM);
2733 		}
2734 
2735 	if (!(flags & FWRITE))
2736 		switch (cmd) {
2737 		case DIOCGETRULES:
2738 		case DIOCGETADDRS:
2739 		case DIOCGETADDR:
2740 		case DIOCGETSTATE:
2741 		case DIOCGETSTATENV:
2742 		case DIOCGETSTATUSNV:
2743 #ifdef COMPAT_FREEBSD14
2744 		case DIOCGETSTATES:
2745 		case DIOCGETSTATESV2:
2746 #endif
2747 		case DIOCGETTIMEOUT:
2748 		case DIOCGETLIMIT:
2749 		case DIOCGETALTQSV0:
2750 		case DIOCGETALTQSV1:
2751 		case DIOCGETALTQV0:
2752 		case DIOCGETALTQV1:
2753 		case DIOCGETQSTATSV0:
2754 		case DIOCGETQSTATSV1:
2755 		case DIOCGETRULESETS:
2756 		case DIOCGETRULESET:
2757 		case DIOCNATLOOK:
2758 		case DIOCRGETTABLES:
2759 		case DIOCRGETTSTATS:
2760 		case DIOCRGETADDRS:
2761 		case DIOCRGETASTATS:
2762 		case DIOCRTSTADDRS:
2763 		case DIOCOSFPGET:
2764 		case DIOCGETSRCNODES:
2765 		case DIOCGETSYNCOOKIES:
2766 		case DIOCIGETIFACES:
2767 		case DIOCGIFSPEEDV1:
2768 		case DIOCGIFSPEEDV0:
2769 		case DIOCGETRULENV:
2770 		case DIOCGETETHRULES:
2771 		case DIOCGETETHRULE:
2772 		case DIOCGETETHRULESETS:
2773 		case DIOCGETETHRULESET:
2774 			break;
2775 		case DIOCRCLRTABLES:
2776 		case DIOCRADDTABLES:
2777 		case DIOCRDELTABLES:
2778 		case DIOCRCLRTSTATS:
2779 		case DIOCRCLRADDRS:
2780 		case DIOCRADDADDRS:
2781 		case DIOCRDELADDRS:
2782 		case DIOCRSETADDRS:
2783 		case DIOCRSETTFLAGS:
2784 			if (((struct pfioc_table *)addr)->pfrio_flags &
2785 			    PFR_FLAG_DUMMY) {
2786 				flags |= FWRITE; /* need write lock for dummy */
2787 				break; /* dummy operation ok */
2788 			}
2789 			return (EACCES);
2790 		default:
2791 			return (EACCES);
2792 		}
2793 
2794 	CURVNET_SET(TD_TO_VNET(td));
2795 
2796 	switch (cmd) {
2797 #ifdef COMPAT_FREEBSD14
2798 	case DIOCSTART:
2799 		error = pf_start();
2800 		break;
2801 
2802 	case DIOCSTOP:
2803 		error = pf_stop();
2804 		break;
2805 #endif
2806 
2807 	case DIOCGETETHRULES: {
2808 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
2809 		nvlist_t		*nvl;
2810 		void			*packed;
2811 		struct pf_keth_rule	*tail;
2812 		struct pf_keth_ruleset	*rs;
2813 		u_int32_t		 ticket, nr;
2814 		const char		*anchor = "";
2815 
2816 		nvl = NULL;
2817 		packed = NULL;
2818 
2819 #define	ERROUT(x)	ERROUT_IOCTL(DIOCGETETHRULES_error, x)
2820 
2821 		if (nv->len > pf_ioctl_maxcount)
2822 			ERROUT(ENOMEM);
2823 
2824 		/* Copy the request in */
2825 		packed = malloc(nv->len, M_NVLIST, M_WAITOK);
2826 		error = copyin(nv->data, packed, nv->len);
2827 		if (error)
2828 			ERROUT(error);
2829 
2830 		nvl = nvlist_unpack(packed, nv->len, 0);
2831 		if (nvl == NULL)
2832 			ERROUT(EBADMSG);
2833 
2834 		if (! nvlist_exists_string(nvl, "anchor"))
2835 			ERROUT(EBADMSG);
2836 
2837 		anchor = nvlist_get_string(nvl, "anchor");
2838 
2839 		rs = pf_find_keth_ruleset(anchor);
2840 
2841 		nvlist_destroy(nvl);
2842 		nvl = NULL;
2843 		free(packed, M_NVLIST);
2844 		packed = NULL;
2845 
2846 		if (rs == NULL)
2847 			ERROUT(ENOENT);
2848 
2849 		/* Reply */
2850 		nvl = nvlist_create(0);
2851 		if (nvl == NULL)
2852 			ERROUT(ENOMEM);
2853 
2854 		PF_RULES_RLOCK();
2855 
2856 		ticket = rs->active.ticket;
2857 		tail = TAILQ_LAST(rs->active.rules, pf_keth_ruleq);
2858 		if (tail)
2859 			nr = tail->nr + 1;
2860 		else
2861 			nr = 0;
2862 
2863 		PF_RULES_RUNLOCK();
2864 
2865 		nvlist_add_number(nvl, "ticket", ticket);
2866 		nvlist_add_number(nvl, "nr", nr);
2867 
2868 		packed = nvlist_pack(nvl, &nv->len);
2869 		if (packed == NULL)
2870 			ERROUT(ENOMEM);
2871 
2872 		if (nv->size == 0)
2873 			ERROUT(0);
2874 		else if (nv->size < nv->len)
2875 			ERROUT(ENOSPC);
2876 
2877 		error = copyout(packed, nv->data, nv->len);
2878 
2879 #undef ERROUT
2880 DIOCGETETHRULES_error:
2881 		free(packed, M_NVLIST);
2882 		nvlist_destroy(nvl);
2883 		break;
2884 	}
2885 
2886 	case DIOCGETETHRULE: {
2887 		struct epoch_tracker	 et;
2888 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
2889 		nvlist_t		*nvl = NULL;
2890 		void			*nvlpacked = NULL;
2891 		struct pf_keth_rule	*rule = NULL;
2892 		struct pf_keth_ruleset	*rs;
2893 		u_int32_t		 ticket, nr;
2894 		bool			 clear = false;
2895 		const char		*anchor;
2896 
2897 #define ERROUT(x)	ERROUT_IOCTL(DIOCGETETHRULE_error, x)
2898 
2899 		if (nv->len > pf_ioctl_maxcount)
2900 			ERROUT(ENOMEM);
2901 
2902 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
2903 		error = copyin(nv->data, nvlpacked, nv->len);
2904 		if (error)
2905 			ERROUT(error);
2906 
2907 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
2908 		if (nvl == NULL)
2909 			ERROUT(EBADMSG);
2910 		if (! nvlist_exists_number(nvl, "ticket"))
2911 			ERROUT(EBADMSG);
2912 		ticket = nvlist_get_number(nvl, "ticket");
2913 		if (! nvlist_exists_string(nvl, "anchor"))
2914 			ERROUT(EBADMSG);
2915 		anchor = nvlist_get_string(nvl, "anchor");
2916 
2917 		if (nvlist_exists_bool(nvl, "clear"))
2918 			clear = nvlist_get_bool(nvl, "clear");
2919 
2920 		if (clear && !(flags & FWRITE))
2921 			ERROUT(EACCES);
2922 
2923 		if (! nvlist_exists_number(nvl, "nr"))
2924 			ERROUT(EBADMSG);
2925 		nr = nvlist_get_number(nvl, "nr");
2926 
2927 		PF_RULES_RLOCK();
2928 		rs = pf_find_keth_ruleset(anchor);
2929 		if (rs == NULL) {
2930 			PF_RULES_RUNLOCK();
2931 			ERROUT(ENOENT);
2932 		}
2933 		if (ticket != rs->active.ticket) {
2934 			PF_RULES_RUNLOCK();
2935 			ERROUT(EBUSY);
2936 		}
2937 
2938 		nvlist_destroy(nvl);
2939 		nvl = NULL;
2940 		free(nvlpacked, M_NVLIST);
2941 		nvlpacked = NULL;
2942 
2943 		rule = TAILQ_FIRST(rs->active.rules);
2944 		while ((rule != NULL) && (rule->nr != nr))
2945 			rule = TAILQ_NEXT(rule, entries);
2946 		if (rule == NULL) {
2947 			PF_RULES_RUNLOCK();
2948 			ERROUT(ENOENT);
2949 		}
2950 		/* Make sure rule can't go away. */
2951 		NET_EPOCH_ENTER(et);
2952 		PF_RULES_RUNLOCK();
2953 		nvl = pf_keth_rule_to_nveth_rule(rule);
2954 		if (pf_keth_anchor_nvcopyout(rs, rule, nvl)) {
2955 			NET_EPOCH_EXIT(et);
2956 			ERROUT(EBUSY);
2957 		}
2958 		NET_EPOCH_EXIT(et);
2959 		if (nvl == NULL)
2960 			ERROUT(ENOMEM);
2961 
2962 		nvlpacked = nvlist_pack(nvl, &nv->len);
2963 		if (nvlpacked == NULL)
2964 			ERROUT(ENOMEM);
2965 
2966 		if (nv->size == 0)
2967 			ERROUT(0);
2968 		else if (nv->size < nv->len)
2969 			ERROUT(ENOSPC);
2970 
2971 		error = copyout(nvlpacked, nv->data, nv->len);
2972 		if (error == 0 && clear) {
2973 			counter_u64_zero(rule->evaluations);
2974 			for (int i = 0; i < 2; i++) {
2975 				counter_u64_zero(rule->packets[i]);
2976 				counter_u64_zero(rule->bytes[i]);
2977 			}
2978 		}
2979 
2980 #undef ERROUT
2981 DIOCGETETHRULE_error:
2982 		free(nvlpacked, M_NVLIST);
2983 		nvlist_destroy(nvl);
2984 		break;
2985 	}
2986 
2987 	case DIOCADDETHRULE: {
2988 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
2989 		nvlist_t		*nvl = NULL;
2990 		void			*nvlpacked = NULL;
2991 		struct pf_keth_rule	*rule = NULL, *tail = NULL;
2992 		struct pf_keth_ruleset	*ruleset = NULL;
2993 		struct pfi_kkif		*kif = NULL, *bridge_to_kif = NULL;
2994 		const char		*anchor = "", *anchor_call = "";
2995 
2996 #define ERROUT(x)	ERROUT_IOCTL(DIOCADDETHRULE_error, x)
2997 
2998 		if (nv->len > pf_ioctl_maxcount)
2999 			ERROUT(ENOMEM);
3000 
3001 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3002 		error = copyin(nv->data, nvlpacked, nv->len);
3003 		if (error)
3004 			ERROUT(error);
3005 
3006 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3007 		if (nvl == NULL)
3008 			ERROUT(EBADMSG);
3009 
3010 		if (! nvlist_exists_number(nvl, "ticket"))
3011 			ERROUT(EBADMSG);
3012 
3013 		if (nvlist_exists_string(nvl, "anchor"))
3014 			anchor = nvlist_get_string(nvl, "anchor");
3015 		if (nvlist_exists_string(nvl, "anchor_call"))
3016 			anchor_call = nvlist_get_string(nvl, "anchor_call");
3017 
3018 		ruleset = pf_find_keth_ruleset(anchor);
3019 		if (ruleset == NULL)
3020 			ERROUT(EINVAL);
3021 
3022 		if (nvlist_get_number(nvl, "ticket") !=
3023 		    ruleset->inactive.ticket) {
3024 			DPFPRINTF(PF_DEBUG_MISC,
3025 			    ("ticket: %d != %d\n",
3026 			    (u_int32_t)nvlist_get_number(nvl, "ticket"),
3027 			    ruleset->inactive.ticket));
3028 			ERROUT(EBUSY);
3029 		}
3030 
3031 		rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
3032 		rule->timestamp = NULL;
3033 
3034 		error = pf_nveth_rule_to_keth_rule(nvl, rule);
3035 		if (error != 0)
3036 			ERROUT(error);
3037 
3038 		if (rule->ifname[0])
3039 			kif = pf_kkif_create(M_WAITOK);
3040 		if (rule->bridge_to_name[0])
3041 			bridge_to_kif = pf_kkif_create(M_WAITOK);
3042 		rule->evaluations = counter_u64_alloc(M_WAITOK);
3043 		for (int i = 0; i < 2; i++) {
3044 			rule->packets[i] = counter_u64_alloc(M_WAITOK);
3045 			rule->bytes[i] = counter_u64_alloc(M_WAITOK);
3046 		}
3047 		rule->timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
3048 		    M_WAITOK | M_ZERO);
3049 
3050 		PF_RULES_WLOCK();
3051 
3052 		if (rule->ifname[0]) {
3053 			rule->kif = pfi_kkif_attach(kif, rule->ifname);
3054 			pfi_kkif_ref(rule->kif);
3055 		} else
3056 			rule->kif = NULL;
3057 		if (rule->bridge_to_name[0]) {
3058 			rule->bridge_to = pfi_kkif_attach(bridge_to_kif,
3059 			    rule->bridge_to_name);
3060 			pfi_kkif_ref(rule->bridge_to);
3061 		} else
3062 			rule->bridge_to = NULL;
3063 
3064 #ifdef ALTQ
3065 		/* set queue IDs */
3066 		if (rule->qname[0] != 0) {
3067 			if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
3068 				error = EBUSY;
3069 			else
3070 				rule->qid = rule->qid;
3071 		}
3072 #endif
3073 		if (rule->tagname[0])
3074 			if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
3075 				error = EBUSY;
3076 		if (rule->match_tagname[0])
3077 			if ((rule->match_tag = pf_tagname2tag(
3078 			    rule->match_tagname)) == 0)
3079 				error = EBUSY;
3080 
3081 		if (error == 0 && rule->ipdst.addr.type == PF_ADDR_TABLE)
3082 			error = pf_eth_addr_setup(ruleset, &rule->ipdst.addr);
3083 		if (error == 0 && rule->ipsrc.addr.type == PF_ADDR_TABLE)
3084 			error = pf_eth_addr_setup(ruleset, &rule->ipsrc.addr);
3085 
3086 		if (error) {
3087 			pf_free_eth_rule(rule);
3088 			PF_RULES_WUNLOCK();
3089 			ERROUT(error);
3090 		}
3091 
3092 		if (pf_keth_anchor_setup(rule, ruleset, anchor_call)) {
3093 			pf_free_eth_rule(rule);
3094 			PF_RULES_WUNLOCK();
3095 			ERROUT(EINVAL);
3096 		}
3097 
3098 		tail = TAILQ_LAST(ruleset->inactive.rules, pf_keth_ruleq);
3099 		if (tail)
3100 			rule->nr = tail->nr + 1;
3101 		else
3102 			rule->nr = 0;
3103 
3104 		TAILQ_INSERT_TAIL(ruleset->inactive.rules, rule, entries);
3105 
3106 		PF_RULES_WUNLOCK();
3107 
3108 #undef ERROUT
3109 DIOCADDETHRULE_error:
3110 		nvlist_destroy(nvl);
3111 		free(nvlpacked, M_NVLIST);
3112 		break;
3113 	}
3114 
3115 	case DIOCGETETHRULESETS: {
3116 		struct epoch_tracker	 et;
3117 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
3118 		nvlist_t		*nvl = NULL;
3119 		void			*nvlpacked = NULL;
3120 		struct pf_keth_ruleset	*ruleset;
3121 		struct pf_keth_anchor	*anchor;
3122 		int			 nr = 0;
3123 
3124 #define ERROUT(x)	ERROUT_IOCTL(DIOCGETETHRULESETS_error, x)
3125 
3126 		if (nv->len > pf_ioctl_maxcount)
3127 			ERROUT(ENOMEM);
3128 
3129 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3130 		error = copyin(nv->data, nvlpacked, nv->len);
3131 		if (error)
3132 			ERROUT(error);
3133 
3134 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3135 		if (nvl == NULL)
3136 			ERROUT(EBADMSG);
3137 		if (! nvlist_exists_string(nvl, "path"))
3138 			ERROUT(EBADMSG);
3139 
3140 		NET_EPOCH_ENTER(et);
3141 
3142 		if ((ruleset = pf_find_keth_ruleset(
3143 		    nvlist_get_string(nvl, "path"))) == NULL) {
3144 			NET_EPOCH_EXIT(et);
3145 			ERROUT(ENOENT);
3146 		}
3147 
3148 		if (ruleset->anchor == NULL) {
3149 			RB_FOREACH(anchor, pf_keth_anchor_global, &V_pf_keth_anchors)
3150 				if (anchor->parent == NULL)
3151 					nr++;
3152 		} else {
3153 			RB_FOREACH(anchor, pf_keth_anchor_node,
3154 			    &ruleset->anchor->children)
3155 				nr++;
3156 		}
3157 
3158 		NET_EPOCH_EXIT(et);
3159 
3160 		nvlist_destroy(nvl);
3161 		nvl = NULL;
3162 		free(nvlpacked, M_NVLIST);
3163 		nvlpacked = NULL;
3164 
3165 		nvl = nvlist_create(0);
3166 		if (nvl == NULL)
3167 			ERROUT(ENOMEM);
3168 
3169 		nvlist_add_number(nvl, "nr", nr);
3170 
3171 		nvlpacked = nvlist_pack(nvl, &nv->len);
3172 		if (nvlpacked == NULL)
3173 			ERROUT(ENOMEM);
3174 
3175 		if (nv->size == 0)
3176 			ERROUT(0);
3177 		else if (nv->size < nv->len)
3178 			ERROUT(ENOSPC);
3179 
3180 		error = copyout(nvlpacked, nv->data, nv->len);
3181 
3182 #undef ERROUT
3183 DIOCGETETHRULESETS_error:
3184 		free(nvlpacked, M_NVLIST);
3185 		nvlist_destroy(nvl);
3186 		break;
3187 	}
3188 
3189 	case DIOCGETETHRULESET: {
3190 		struct epoch_tracker	 et;
3191 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
3192 		nvlist_t		*nvl = NULL;
3193 		void			*nvlpacked = NULL;
3194 		struct pf_keth_ruleset	*ruleset;
3195 		struct pf_keth_anchor	*anchor;
3196 		int			 nr = 0, req_nr = 0;
3197 		bool			 found = false;
3198 
3199 #define ERROUT(x)	ERROUT_IOCTL(DIOCGETETHRULESET_error, x)
3200 
3201 		if (nv->len > pf_ioctl_maxcount)
3202 			ERROUT(ENOMEM);
3203 
3204 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3205 		error = copyin(nv->data, nvlpacked, nv->len);
3206 		if (error)
3207 			ERROUT(error);
3208 
3209 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3210 		if (nvl == NULL)
3211 			ERROUT(EBADMSG);
3212 		if (! nvlist_exists_string(nvl, "path"))
3213 			ERROUT(EBADMSG);
3214 		if (! nvlist_exists_number(nvl, "nr"))
3215 			ERROUT(EBADMSG);
3216 
3217 		req_nr = nvlist_get_number(nvl, "nr");
3218 
3219 		NET_EPOCH_ENTER(et);
3220 
3221 		if ((ruleset = pf_find_keth_ruleset(
3222 		    nvlist_get_string(nvl, "path"))) == NULL) {
3223 			NET_EPOCH_EXIT(et);
3224 			ERROUT(ENOENT);
3225 		}
3226 
3227 		nvlist_destroy(nvl);
3228 		nvl = NULL;
3229 		free(nvlpacked, M_NVLIST);
3230 		nvlpacked = NULL;
3231 
3232 		nvl = nvlist_create(0);
3233 		if (nvl == NULL) {
3234 			NET_EPOCH_EXIT(et);
3235 			ERROUT(ENOMEM);
3236 		}
3237 
3238 		if (ruleset->anchor == NULL) {
3239 			RB_FOREACH(anchor, pf_keth_anchor_global,
3240 			    &V_pf_keth_anchors) {
3241 				if (anchor->parent == NULL && nr++ == req_nr) {
3242 					found = true;
3243 					break;
3244 				}
3245 			}
3246 		} else {
3247 			RB_FOREACH(anchor, pf_keth_anchor_node,
3248 			     &ruleset->anchor->children) {
3249 				if (nr++ == req_nr) {
3250 					found = true;
3251 					break;
3252 				}
3253 			}
3254 		}
3255 
3256 		NET_EPOCH_EXIT(et);
3257 		if (found) {
3258 			nvlist_add_number(nvl, "nr", nr);
3259 			nvlist_add_string(nvl, "name", anchor->name);
3260 			if (ruleset->anchor)
3261 				nvlist_add_string(nvl, "path",
3262 				    ruleset->anchor->path);
3263 			else
3264 				nvlist_add_string(nvl, "path", "");
3265 		} else {
3266 			ERROUT(EBUSY);
3267 		}
3268 
3269 		nvlpacked = nvlist_pack(nvl, &nv->len);
3270 		if (nvlpacked == NULL)
3271 			ERROUT(ENOMEM);
3272 
3273 		if (nv->size == 0)
3274 			ERROUT(0);
3275 		else if (nv->size < nv->len)
3276 			ERROUT(ENOSPC);
3277 
3278 		error = copyout(nvlpacked, nv->data, nv->len);
3279 
3280 #undef ERROUT
3281 DIOCGETETHRULESET_error:
3282 		free(nvlpacked, M_NVLIST);
3283 		nvlist_destroy(nvl);
3284 		break;
3285 	}
3286 
3287 	case DIOCADDRULENV: {
3288 		struct pfioc_nv	*nv = (struct pfioc_nv *)addr;
3289 		nvlist_t	*nvl = NULL;
3290 		void		*nvlpacked = NULL;
3291 		struct pf_krule	*rule = NULL;
3292 		const char	*anchor = "", *anchor_call = "";
3293 		uint32_t	 ticket = 0, pool_ticket = 0;
3294 
3295 #define	ERROUT(x)	ERROUT_IOCTL(DIOCADDRULENV_error, x)
3296 
3297 		if (nv->len > pf_ioctl_maxcount)
3298 			ERROUT(ENOMEM);
3299 
3300 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3301 		error = copyin(nv->data, nvlpacked, nv->len);
3302 		if (error)
3303 			ERROUT(error);
3304 
3305 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3306 		if (nvl == NULL)
3307 			ERROUT(EBADMSG);
3308 
3309 		if (! nvlist_exists_number(nvl, "ticket"))
3310 			ERROUT(EINVAL);
3311 		ticket = nvlist_get_number(nvl, "ticket");
3312 
3313 		if (! nvlist_exists_number(nvl, "pool_ticket"))
3314 			ERROUT(EINVAL);
3315 		pool_ticket = nvlist_get_number(nvl, "pool_ticket");
3316 
3317 		if (! nvlist_exists_nvlist(nvl, "rule"))
3318 			ERROUT(EINVAL);
3319 
3320 		rule = pf_krule_alloc();
3321 		error = pf_nvrule_to_krule(nvlist_get_nvlist(nvl, "rule"),
3322 		    rule);
3323 		if (error)
3324 			ERROUT(error);
3325 
3326 		if (nvlist_exists_string(nvl, "anchor"))
3327 			anchor = nvlist_get_string(nvl, "anchor");
3328 		if (nvlist_exists_string(nvl, "anchor_call"))
3329 			anchor_call = nvlist_get_string(nvl, "anchor_call");
3330 
3331 		if ((error = nvlist_error(nvl)))
3332 			ERROUT(error);
3333 
3334 		/* Frees rule on error */
3335 		error = pf_ioctl_addrule(rule, ticket, pool_ticket, anchor,
3336 		    anchor_call, td->td_ucred->cr_ruid,
3337 		    td->td_proc ? td->td_proc->p_pid : 0);
3338 
3339 		nvlist_destroy(nvl);
3340 		free(nvlpacked, M_NVLIST);
3341 		break;
3342 #undef ERROUT
3343 DIOCADDRULENV_error:
3344 		pf_krule_free(rule);
3345 		nvlist_destroy(nvl);
3346 		free(nvlpacked, M_NVLIST);
3347 
3348 		break;
3349 	}
3350 	case DIOCADDRULE: {
3351 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
3352 		struct pf_krule		*rule;
3353 
3354 		rule = pf_krule_alloc();
3355 		error = pf_rule_to_krule(&pr->rule, rule);
3356 		if (error != 0) {
3357 			pf_krule_free(rule);
3358 			break;
3359 		}
3360 
3361 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
3362 
3363 		/* Frees rule on error */
3364 		error = pf_ioctl_addrule(rule, pr->ticket, pr->pool_ticket,
3365 		    pr->anchor, pr->anchor_call, td->td_ucred->cr_ruid,
3366 		    td->td_proc ? td->td_proc->p_pid : 0);
3367 		break;
3368 	}
3369 
3370 	case DIOCGETRULES: {
3371 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
3372 
3373 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
3374 
3375 		error = pf_ioctl_getrules(pr);
3376 
3377 		break;
3378 	}
3379 
3380 	case DIOCGETRULENV: {
3381 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
3382 		nvlist_t		*nvrule = NULL;
3383 		nvlist_t		*nvl = NULL;
3384 		struct pf_kruleset	*ruleset;
3385 		struct pf_krule		*rule;
3386 		void			*nvlpacked = NULL;
3387 		int			 rs_num, nr;
3388 		bool			 clear_counter = false;
3389 
3390 #define	ERROUT(x)	ERROUT_IOCTL(DIOCGETRULENV_error, x)
3391 
3392 		if (nv->len > pf_ioctl_maxcount)
3393 			ERROUT(ENOMEM);
3394 
3395 		/* Copy the request in */
3396 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3397 		error = copyin(nv->data, nvlpacked, nv->len);
3398 		if (error)
3399 			ERROUT(error);
3400 
3401 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3402 		if (nvl == NULL)
3403 			ERROUT(EBADMSG);
3404 
3405 		if (! nvlist_exists_string(nvl, "anchor"))
3406 			ERROUT(EBADMSG);
3407 		if (! nvlist_exists_number(nvl, "ruleset"))
3408 			ERROUT(EBADMSG);
3409 		if (! nvlist_exists_number(nvl, "ticket"))
3410 			ERROUT(EBADMSG);
3411 		if (! nvlist_exists_number(nvl, "nr"))
3412 			ERROUT(EBADMSG);
3413 
3414 		if (nvlist_exists_bool(nvl, "clear_counter"))
3415 			clear_counter = nvlist_get_bool(nvl, "clear_counter");
3416 
3417 		if (clear_counter && !(flags & FWRITE))
3418 			ERROUT(EACCES);
3419 
3420 		nr = nvlist_get_number(nvl, "nr");
3421 
3422 		PF_RULES_WLOCK();
3423 		ruleset = pf_find_kruleset(nvlist_get_string(nvl, "anchor"));
3424 		if (ruleset == NULL) {
3425 			PF_RULES_WUNLOCK();
3426 			ERROUT(ENOENT);
3427 		}
3428 
3429 		rs_num = pf_get_ruleset_number(nvlist_get_number(nvl, "ruleset"));
3430 		if (rs_num >= PF_RULESET_MAX) {
3431 			PF_RULES_WUNLOCK();
3432 			ERROUT(EINVAL);
3433 		}
3434 
3435 		if (nvlist_get_number(nvl, "ticket") !=
3436 		    ruleset->rules[rs_num].active.ticket) {
3437 			PF_RULES_WUNLOCK();
3438 			ERROUT(EBUSY);
3439 		}
3440 
3441 		if ((error = nvlist_error(nvl))) {
3442 			PF_RULES_WUNLOCK();
3443 			ERROUT(error);
3444 		}
3445 
3446 		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
3447 		while ((rule != NULL) && (rule->nr != nr))
3448 			rule = TAILQ_NEXT(rule, entries);
3449 		if (rule == NULL) {
3450 			PF_RULES_WUNLOCK();
3451 			ERROUT(EBUSY);
3452 		}
3453 
3454 		nvrule = pf_krule_to_nvrule(rule);
3455 
3456 		nvlist_destroy(nvl);
3457 		nvl = nvlist_create(0);
3458 		if (nvl == NULL) {
3459 			PF_RULES_WUNLOCK();
3460 			ERROUT(ENOMEM);
3461 		}
3462 		nvlist_add_number(nvl, "nr", nr);
3463 		nvlist_add_nvlist(nvl, "rule", nvrule);
3464 		nvlist_destroy(nvrule);
3465 		nvrule = NULL;
3466 		if (pf_kanchor_nvcopyout(ruleset, rule, nvl)) {
3467 			PF_RULES_WUNLOCK();
3468 			ERROUT(EBUSY);
3469 		}
3470 
3471 		free(nvlpacked, M_NVLIST);
3472 		nvlpacked = nvlist_pack(nvl, &nv->len);
3473 		if (nvlpacked == NULL) {
3474 			PF_RULES_WUNLOCK();
3475 			ERROUT(ENOMEM);
3476 		}
3477 
3478 		if (nv->size == 0) {
3479 			PF_RULES_WUNLOCK();
3480 			ERROUT(0);
3481 		}
3482 		else if (nv->size < nv->len) {
3483 			PF_RULES_WUNLOCK();
3484 			ERROUT(ENOSPC);
3485 		}
3486 
3487 		if (clear_counter)
3488 			pf_krule_clear_counters(rule);
3489 
3490 		PF_RULES_WUNLOCK();
3491 
3492 		error = copyout(nvlpacked, nv->data, nv->len);
3493 
3494 #undef ERROUT
3495 DIOCGETRULENV_error:
3496 		free(nvlpacked, M_NVLIST);
3497 		nvlist_destroy(nvrule);
3498 		nvlist_destroy(nvl);
3499 
3500 		break;
3501 	}
3502 
3503 	case DIOCCHANGERULE: {
3504 		struct pfioc_rule	*pcr = (struct pfioc_rule *)addr;
3505 		struct pf_kruleset	*ruleset;
3506 		struct pf_krule		*oldrule = NULL, *newrule = NULL;
3507 		struct pfi_kkif		*kif = NULL;
3508 		struct pf_kpooladdr	*pa;
3509 		u_int32_t		 nr = 0;
3510 		int			 rs_num;
3511 
3512 		pcr->anchor[sizeof(pcr->anchor) - 1] = 0;
3513 
3514 		if (pcr->action < PF_CHANGE_ADD_HEAD ||
3515 		    pcr->action > PF_CHANGE_GET_TICKET) {
3516 			error = EINVAL;
3517 			break;
3518 		}
3519 		if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
3520 			error = EINVAL;
3521 			break;
3522 		}
3523 
3524 		if (pcr->action != PF_CHANGE_REMOVE) {
3525 			newrule = pf_krule_alloc();
3526 			error = pf_rule_to_krule(&pcr->rule, newrule);
3527 			if (error != 0) {
3528 				pf_krule_free(newrule);
3529 				break;
3530 			}
3531 
3532 			if (newrule->ifname[0])
3533 				kif = pf_kkif_create(M_WAITOK);
3534 			pf_counter_u64_init(&newrule->evaluations, M_WAITOK);
3535 			for (int i = 0; i < 2; i++) {
3536 				pf_counter_u64_init(&newrule->packets[i], M_WAITOK);
3537 				pf_counter_u64_init(&newrule->bytes[i], M_WAITOK);
3538 			}
3539 			newrule->states_cur = counter_u64_alloc(M_WAITOK);
3540 			newrule->states_tot = counter_u64_alloc(M_WAITOK);
3541 			newrule->src_nodes = counter_u64_alloc(M_WAITOK);
3542 			newrule->cuid = td->td_ucred->cr_ruid;
3543 			newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
3544 			TAILQ_INIT(&newrule->rpool.list);
3545 		}
3546 #define	ERROUT(x)	ERROUT_IOCTL(DIOCCHANGERULE_error, x)
3547 
3548 		PF_CONFIG_LOCK();
3549 		PF_RULES_WLOCK();
3550 #ifdef PF_WANT_32_TO_64_COUNTER
3551 		if (newrule != NULL) {
3552 			LIST_INSERT_HEAD(&V_pf_allrulelist, newrule, allrulelist);
3553 			newrule->allrulelinked = true;
3554 			V_pf_allrulecount++;
3555 		}
3556 #endif
3557 
3558 		if (!(pcr->action == PF_CHANGE_REMOVE ||
3559 		    pcr->action == PF_CHANGE_GET_TICKET) &&
3560 		    pcr->pool_ticket != V_ticket_pabuf)
3561 			ERROUT(EBUSY);
3562 
3563 		ruleset = pf_find_kruleset(pcr->anchor);
3564 		if (ruleset == NULL)
3565 			ERROUT(EINVAL);
3566 
3567 		rs_num = pf_get_ruleset_number(pcr->rule.action);
3568 		if (rs_num >= PF_RULESET_MAX)
3569 			ERROUT(EINVAL);
3570 
3571 		/*
3572 		 * XXXMJG: there is no guarantee that the ruleset was
3573 		 * created by the usual route of calling DIOCXBEGIN.
3574 		 * As a result it is possible the rule tree will not
3575 		 * be allocated yet. Hack around it by doing it here.
3576 		 * Note it is fine to let the tree persist in case of
3577 		 * error as it will be freed down the road on future
3578 		 * updates (if need be).
3579 		 */
3580 		if (ruleset->rules[rs_num].active.tree == NULL) {
3581 			ruleset->rules[rs_num].active.tree = pf_rule_tree_alloc(M_NOWAIT);
3582 			if (ruleset->rules[rs_num].active.tree == NULL) {
3583 				ERROUT(ENOMEM);
3584 			}
3585 		}
3586 
3587 		if (pcr->action == PF_CHANGE_GET_TICKET) {
3588 			pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
3589 			ERROUT(0);
3590 		} else if (pcr->ticket !=
3591 			    ruleset->rules[rs_num].active.ticket)
3592 				ERROUT(EINVAL);
3593 
3594 		if (pcr->action != PF_CHANGE_REMOVE) {
3595 			if (newrule->ifname[0]) {
3596 				newrule->kif = pfi_kkif_attach(kif,
3597 				    newrule->ifname);
3598 				kif = NULL;
3599 				pfi_kkif_ref(newrule->kif);
3600 			} else
3601 				newrule->kif = NULL;
3602 
3603 			if (newrule->rtableid > 0 &&
3604 			    newrule->rtableid >= rt_numfibs)
3605 				error = EBUSY;
3606 
3607 #ifdef ALTQ
3608 			/* set queue IDs */
3609 			if (newrule->qname[0] != 0) {
3610 				if ((newrule->qid =
3611 				    pf_qname2qid(newrule->qname)) == 0)
3612 					error = EBUSY;
3613 				else if (newrule->pqname[0] != 0) {
3614 					if ((newrule->pqid =
3615 					    pf_qname2qid(newrule->pqname)) == 0)
3616 						error = EBUSY;
3617 				} else
3618 					newrule->pqid = newrule->qid;
3619 			}
3620 #endif /* ALTQ */
3621 			if (newrule->tagname[0])
3622 				if ((newrule->tag =
3623 				    pf_tagname2tag(newrule->tagname)) == 0)
3624 					error = EBUSY;
3625 			if (newrule->match_tagname[0])
3626 				if ((newrule->match_tag = pf_tagname2tag(
3627 				    newrule->match_tagname)) == 0)
3628 					error = EBUSY;
3629 			if (newrule->rt && !newrule->direction)
3630 				error = EINVAL;
3631 			if (!newrule->log)
3632 				newrule->logif = 0;
3633 			if (newrule->logif >= PFLOGIFS_MAX)
3634 				error = EINVAL;
3635 			if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
3636 				error = ENOMEM;
3637 			if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
3638 				error = ENOMEM;
3639 			if (pf_kanchor_setup(newrule, ruleset, pcr->anchor_call))
3640 				error = EINVAL;
3641 			TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
3642 				if (pa->addr.type == PF_ADDR_TABLE) {
3643 					pa->addr.p.tbl =
3644 					    pfr_attach_table(ruleset,
3645 					    pa->addr.v.tblname);
3646 					if (pa->addr.p.tbl == NULL)
3647 						error = ENOMEM;
3648 				}
3649 
3650 			newrule->overload_tbl = NULL;
3651 			if (newrule->overload_tblname[0]) {
3652 				if ((newrule->overload_tbl = pfr_attach_table(
3653 				    ruleset, newrule->overload_tblname)) ==
3654 				    NULL)
3655 					error = EINVAL;
3656 				else
3657 					newrule->overload_tbl->pfrkt_flags |=
3658 					    PFR_TFLAG_ACTIVE;
3659 			}
3660 
3661 			pf_mv_kpool(&V_pf_pabuf, &newrule->rpool.list);
3662 			if (((((newrule->action == PF_NAT) ||
3663 			    (newrule->action == PF_RDR) ||
3664 			    (newrule->action == PF_BINAT) ||
3665 			    (newrule->rt > PF_NOPFROUTE)) &&
3666 			    !newrule->anchor)) &&
3667 			    (TAILQ_FIRST(&newrule->rpool.list) == NULL))
3668 				error = EINVAL;
3669 
3670 			if (error) {
3671 				pf_free_rule(newrule);
3672 				PF_RULES_WUNLOCK();
3673 				PF_CONFIG_UNLOCK();
3674 				break;
3675 			}
3676 
3677 			newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
3678 		}
3679 		pf_empty_kpool(&V_pf_pabuf);
3680 
3681 		if (pcr->action == PF_CHANGE_ADD_HEAD)
3682 			oldrule = TAILQ_FIRST(
3683 			    ruleset->rules[rs_num].active.ptr);
3684 		else if (pcr->action == PF_CHANGE_ADD_TAIL)
3685 			oldrule = TAILQ_LAST(
3686 			    ruleset->rules[rs_num].active.ptr, pf_krulequeue);
3687 		else {
3688 			oldrule = TAILQ_FIRST(
3689 			    ruleset->rules[rs_num].active.ptr);
3690 			while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
3691 				oldrule = TAILQ_NEXT(oldrule, entries);
3692 			if (oldrule == NULL) {
3693 				if (newrule != NULL)
3694 					pf_free_rule(newrule);
3695 				PF_RULES_WUNLOCK();
3696 				PF_CONFIG_UNLOCK();
3697 				error = EINVAL;
3698 				break;
3699 			}
3700 		}
3701 
3702 		if (pcr->action == PF_CHANGE_REMOVE) {
3703 			pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
3704 			    oldrule);
3705 			RB_REMOVE(pf_krule_global,
3706 			    ruleset->rules[rs_num].active.tree, oldrule);
3707 			ruleset->rules[rs_num].active.rcount--;
3708 		} else {
3709 			pf_hash_rule(newrule);
3710 			if (RB_INSERT(pf_krule_global,
3711 			    ruleset->rules[rs_num].active.tree, newrule) != NULL) {
3712 				pf_free_rule(newrule);
3713 				PF_RULES_WUNLOCK();
3714 				PF_CONFIG_UNLOCK();
3715 				error = EEXIST;
3716 				break;
3717 			}
3718 
3719 			if (oldrule == NULL)
3720 				TAILQ_INSERT_TAIL(
3721 				    ruleset->rules[rs_num].active.ptr,
3722 				    newrule, entries);
3723 			else if (pcr->action == PF_CHANGE_ADD_HEAD ||
3724 			    pcr->action == PF_CHANGE_ADD_BEFORE)
3725 				TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
3726 			else
3727 				TAILQ_INSERT_AFTER(
3728 				    ruleset->rules[rs_num].active.ptr,
3729 				    oldrule, newrule, entries);
3730 			ruleset->rules[rs_num].active.rcount++;
3731 		}
3732 
3733 		nr = 0;
3734 		TAILQ_FOREACH(oldrule,
3735 		    ruleset->rules[rs_num].active.ptr, entries)
3736 			oldrule->nr = nr++;
3737 
3738 		ruleset->rules[rs_num].active.ticket++;
3739 
3740 		pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
3741 		pf_remove_if_empty_kruleset(ruleset);
3742 
3743 		PF_RULES_WUNLOCK();
3744 		PF_CONFIG_UNLOCK();
3745 		break;
3746 
3747 #undef ERROUT
3748 DIOCCHANGERULE_error:
3749 		PF_RULES_WUNLOCK();
3750 		PF_CONFIG_UNLOCK();
3751 		pf_krule_free(newrule);
3752 		pf_kkif_free(kif);
3753 		break;
3754 	}
3755 
3756 	case DIOCCLRSTATESNV: {
3757 		error = pf_clearstates_nv((struct pfioc_nv *)addr);
3758 		break;
3759 	}
3760 
3761 	case DIOCKILLSTATESNV: {
3762 		error = pf_killstates_nv((struct pfioc_nv *)addr);
3763 		break;
3764 	}
3765 
3766 	case DIOCADDSTATE: {
3767 		struct pfioc_state		*ps = (struct pfioc_state *)addr;
3768 		struct pfsync_state_1301	*sp = &ps->state;
3769 
3770 		if (sp->timeout >= PFTM_MAX) {
3771 			error = EINVAL;
3772 			break;
3773 		}
3774 		if (V_pfsync_state_import_ptr != NULL) {
3775 			PF_RULES_RLOCK();
3776 			error = V_pfsync_state_import_ptr(
3777 			    (union pfsync_state_union *)sp, PFSYNC_SI_IOCTL,
3778 			    PFSYNC_MSG_VERSION_1301);
3779 			PF_RULES_RUNLOCK();
3780 		} else
3781 			error = EOPNOTSUPP;
3782 		break;
3783 	}
3784 
3785 	case DIOCGETSTATE: {
3786 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
3787 		struct pf_kstate	*s;
3788 
3789 		s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
3790 		if (s == NULL) {
3791 			error = ENOENT;
3792 			break;
3793 		}
3794 
3795 		pfsync_state_export((union pfsync_state_union*)&ps->state,
3796 		    s, PFSYNC_MSG_VERSION_1301);
3797 		PF_STATE_UNLOCK(s);
3798 		break;
3799 	}
3800 
3801 	case DIOCGETSTATENV: {
3802 		error = pf_getstate((struct pfioc_nv *)addr);
3803 		break;
3804 	}
3805 
3806 #ifdef COMPAT_FREEBSD14
3807 	case DIOCGETSTATES: {
3808 		struct pfioc_states	*ps = (struct pfioc_states *)addr;
3809 		struct pf_kstate	*s;
3810 		struct pfsync_state_1301	*pstore, *p;
3811 		int			 i, nr;
3812 		size_t			 slice_count = 16, count;
3813 		void			*out;
3814 
3815 		if (ps->ps_len <= 0) {
3816 			nr = uma_zone_get_cur(V_pf_state_z);
3817 			ps->ps_len = sizeof(struct pfsync_state_1301) * nr;
3818 			break;
3819 		}
3820 
3821 		out = ps->ps_states;
3822 		pstore = mallocarray(slice_count,
3823 		    sizeof(struct pfsync_state_1301), M_TEMP, M_WAITOK | M_ZERO);
3824 		nr = 0;
3825 
3826 		for (i = 0; i <= V_pf_hashmask; i++) {
3827 			struct pf_idhash *ih = &V_pf_idhash[i];
3828 
3829 DIOCGETSTATES_retry:
3830 			p = pstore;
3831 
3832 			if (LIST_EMPTY(&ih->states))
3833 				continue;
3834 
3835 			PF_HASHROW_LOCK(ih);
3836 			count = 0;
3837 			LIST_FOREACH(s, &ih->states, entry) {
3838 				if (s->timeout == PFTM_UNLINKED)
3839 					continue;
3840 				count++;
3841 			}
3842 
3843 			if (count > slice_count) {
3844 				PF_HASHROW_UNLOCK(ih);
3845 				free(pstore, M_TEMP);
3846 				slice_count = count * 2;
3847 				pstore = mallocarray(slice_count,
3848 				    sizeof(struct pfsync_state_1301), M_TEMP,
3849 				    M_WAITOK | M_ZERO);
3850 				goto DIOCGETSTATES_retry;
3851 			}
3852 
3853 			if ((nr+count) * sizeof(*p) > ps->ps_len) {
3854 				PF_HASHROW_UNLOCK(ih);
3855 				goto DIOCGETSTATES_full;
3856 			}
3857 
3858 			LIST_FOREACH(s, &ih->states, entry) {
3859 				if (s->timeout == PFTM_UNLINKED)
3860 					continue;
3861 
3862 				pfsync_state_export((union pfsync_state_union*)p,
3863 				    s, PFSYNC_MSG_VERSION_1301);
3864 				p++;
3865 				nr++;
3866 			}
3867 			PF_HASHROW_UNLOCK(ih);
3868 			error = copyout(pstore, out,
3869 			    sizeof(struct pfsync_state_1301) * count);
3870 			if (error)
3871 				break;
3872 			out = ps->ps_states + nr;
3873 		}
3874 DIOCGETSTATES_full:
3875 		ps->ps_len = sizeof(struct pfsync_state_1301) * nr;
3876 		free(pstore, M_TEMP);
3877 
3878 		break;
3879 	}
3880 
3881 	case DIOCGETSTATESV2: {
3882 		struct pfioc_states_v2	*ps = (struct pfioc_states_v2 *)addr;
3883 		struct pf_kstate	*s;
3884 		struct pf_state_export	*pstore, *p;
3885 		int i, nr;
3886 		size_t slice_count = 16, count;
3887 		void *out;
3888 
3889 		if (ps->ps_req_version > PF_STATE_VERSION) {
3890 			error = ENOTSUP;
3891 			break;
3892 		}
3893 
3894 		if (ps->ps_len <= 0) {
3895 			nr = uma_zone_get_cur(V_pf_state_z);
3896 			ps->ps_len = sizeof(struct pf_state_export) * nr;
3897 			break;
3898 		}
3899 
3900 		out = ps->ps_states;
3901 		pstore = mallocarray(slice_count,
3902 		    sizeof(struct pf_state_export), M_TEMP, M_WAITOK | M_ZERO);
3903 		nr = 0;
3904 
3905 		for (i = 0; i <= V_pf_hashmask; i++) {
3906 			struct pf_idhash *ih = &V_pf_idhash[i];
3907 
3908 DIOCGETSTATESV2_retry:
3909 			p = pstore;
3910 
3911 			if (LIST_EMPTY(&ih->states))
3912 				continue;
3913 
3914 			PF_HASHROW_LOCK(ih);
3915 			count = 0;
3916 			LIST_FOREACH(s, &ih->states, entry) {
3917 				if (s->timeout == PFTM_UNLINKED)
3918 					continue;
3919 				count++;
3920 			}
3921 
3922 			if (count > slice_count) {
3923 				PF_HASHROW_UNLOCK(ih);
3924 				free(pstore, M_TEMP);
3925 				slice_count = count * 2;
3926 				pstore = mallocarray(slice_count,
3927 				    sizeof(struct pf_state_export), M_TEMP,
3928 				    M_WAITOK | M_ZERO);
3929 				goto DIOCGETSTATESV2_retry;
3930 			}
3931 
3932 			if ((nr+count) * sizeof(*p) > ps->ps_len) {
3933 				PF_HASHROW_UNLOCK(ih);
3934 				goto DIOCGETSTATESV2_full;
3935 			}
3936 
3937 			LIST_FOREACH(s, &ih->states, entry) {
3938 				if (s->timeout == PFTM_UNLINKED)
3939 					continue;
3940 
3941 				pf_state_export(p, s);
3942 				p++;
3943 				nr++;
3944 			}
3945 			PF_HASHROW_UNLOCK(ih);
3946 			error = copyout(pstore, out,
3947 			    sizeof(struct pf_state_export) * count);
3948 			if (error)
3949 				break;
3950 			out = ps->ps_states + nr;
3951 		}
3952 DIOCGETSTATESV2_full:
3953 		ps->ps_len = nr * sizeof(struct pf_state_export);
3954 		free(pstore, M_TEMP);
3955 
3956 		break;
3957 	}
3958 #endif
3959 	case DIOCGETSTATUSNV: {
3960 		error = pf_getstatus((struct pfioc_nv *)addr);
3961 		break;
3962 	}
3963 
3964 	case DIOCSETSTATUSIF: {
3965 		struct pfioc_if	*pi = (struct pfioc_if *)addr;
3966 
3967 		if (pi->ifname[0] == 0) {
3968 			bzero(V_pf_status.ifname, IFNAMSIZ);
3969 			break;
3970 		}
3971 		PF_RULES_WLOCK();
3972 		error = pf_user_strcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
3973 		PF_RULES_WUNLOCK();
3974 		break;
3975 	}
3976 
3977 	case DIOCCLRSTATUS: {
3978 		pf_ioctl_clear_status();
3979 		break;
3980 	}
3981 
3982 	case DIOCNATLOOK: {
3983 		struct pfioc_natlook	*pnl = (struct pfioc_natlook *)addr;
3984 		struct pf_state_key	*sk;
3985 		struct pf_kstate	*state;
3986 		struct pf_state_key_cmp	 key;
3987 		int			 m = 0, direction = pnl->direction;
3988 		int			 sidx, didx;
3989 
3990 		/* NATLOOK src and dst are reversed, so reverse sidx/didx */
3991 		sidx = (direction == PF_IN) ? 1 : 0;
3992 		didx = (direction == PF_IN) ? 0 : 1;
3993 
3994 		if (!pnl->proto ||
3995 		    PF_AZERO(&pnl->saddr, pnl->af) ||
3996 		    PF_AZERO(&pnl->daddr, pnl->af) ||
3997 		    ((pnl->proto == IPPROTO_TCP ||
3998 		    pnl->proto == IPPROTO_UDP) &&
3999 		    (!pnl->dport || !pnl->sport)))
4000 			error = EINVAL;
4001 		else {
4002 			bzero(&key, sizeof(key));
4003 			key.af = pnl->af;
4004 			key.proto = pnl->proto;
4005 			PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
4006 			key.port[sidx] = pnl->sport;
4007 			PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
4008 			key.port[didx] = pnl->dport;
4009 
4010 			state = pf_find_state_all(&key, direction, &m);
4011 			if (state == NULL) {
4012 				error = ENOENT;
4013 			} else {
4014 				if (m > 1) {
4015 					PF_STATE_UNLOCK(state);
4016 					error = E2BIG;	/* more than one state */
4017 				} else {
4018 					sk = state->key[sidx];
4019 					PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
4020 					pnl->rsport = sk->port[sidx];
4021 					PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
4022 					pnl->rdport = sk->port[didx];
4023 					PF_STATE_UNLOCK(state);
4024 				}
4025 			}
4026 		}
4027 		break;
4028 	}
4029 
4030 	case DIOCSETTIMEOUT: {
4031 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
4032 
4033 		error = pf_ioctl_set_timeout(pt->timeout, pt->seconds,
4034 		    &pt->seconds);
4035 		break;
4036 	}
4037 
4038 	case DIOCGETTIMEOUT: {
4039 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
4040 
4041 		error = pf_ioctl_get_timeout(pt->timeout, &pt->seconds);
4042 		break;
4043 	}
4044 
4045 	case DIOCGETLIMIT: {
4046 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
4047 
4048 		error = pf_ioctl_get_limit(pl->index, &pl->limit);
4049 		break;
4050 	}
4051 
4052 	case DIOCSETLIMIT: {
4053 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
4054 		unsigned int old_limit;
4055 
4056 		error = pf_ioctl_set_limit(pl->index, pl->limit, &old_limit);
4057 		pl->limit = old_limit;
4058 		break;
4059 	}
4060 
4061 	case DIOCSETDEBUG: {
4062 		u_int32_t	*level = (u_int32_t *)addr;
4063 
4064 		PF_RULES_WLOCK();
4065 		V_pf_status.debug = *level;
4066 		PF_RULES_WUNLOCK();
4067 		break;
4068 	}
4069 
4070 	case DIOCCLRRULECTRS: {
4071 		/* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
4072 		struct pf_kruleset	*ruleset = &pf_main_ruleset;
4073 		struct pf_krule		*rule;
4074 
4075 		PF_RULES_WLOCK();
4076 		TAILQ_FOREACH(rule,
4077 		    ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
4078 			pf_counter_u64_zero(&rule->evaluations);
4079 			for (int i = 0; i < 2; i++) {
4080 				pf_counter_u64_zero(&rule->packets[i]);
4081 				pf_counter_u64_zero(&rule->bytes[i]);
4082 			}
4083 		}
4084 		PF_RULES_WUNLOCK();
4085 		break;
4086 	}
4087 
4088 	case DIOCGIFSPEEDV0:
4089 	case DIOCGIFSPEEDV1: {
4090 		struct pf_ifspeed_v1	*psp = (struct pf_ifspeed_v1 *)addr;
4091 		struct pf_ifspeed_v1	ps;
4092 		struct ifnet		*ifp;
4093 
4094 		if (psp->ifname[0] == '\0') {
4095 			error = EINVAL;
4096 			break;
4097 		}
4098 
4099 		error = pf_user_strcpy(ps.ifname, psp->ifname, IFNAMSIZ);
4100 		if (error != 0)
4101 			break;
4102 		ifp = ifunit(ps.ifname);
4103 		if (ifp != NULL) {
4104 			psp->baudrate32 =
4105 			    (u_int32_t)uqmin(ifp->if_baudrate, UINT_MAX);
4106 			if (cmd == DIOCGIFSPEEDV1)
4107 				psp->baudrate = ifp->if_baudrate;
4108 		} else {
4109 			error = EINVAL;
4110 		}
4111 		break;
4112 	}
4113 
4114 #ifdef ALTQ
4115 	case DIOCSTARTALTQ: {
4116 		struct pf_altq		*altq;
4117 
4118 		PF_RULES_WLOCK();
4119 		/* enable all altq interfaces on active list */
4120 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
4121 			if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
4122 				error = pf_enable_altq(altq);
4123 				if (error != 0)
4124 					break;
4125 			}
4126 		}
4127 		if (error == 0)
4128 			V_pf_altq_running = 1;
4129 		PF_RULES_WUNLOCK();
4130 		DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
4131 		break;
4132 	}
4133 
4134 	case DIOCSTOPALTQ: {
4135 		struct pf_altq		*altq;
4136 
4137 		PF_RULES_WLOCK();
4138 		/* disable all altq interfaces on active list */
4139 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
4140 			if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
4141 				error = pf_disable_altq(altq);
4142 				if (error != 0)
4143 					break;
4144 			}
4145 		}
4146 		if (error == 0)
4147 			V_pf_altq_running = 0;
4148 		PF_RULES_WUNLOCK();
4149 		DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
4150 		break;
4151 	}
4152 
4153 	case DIOCADDALTQV0:
4154 	case DIOCADDALTQV1: {
4155 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
4156 		struct pf_altq		*altq, *a;
4157 		struct ifnet		*ifp;
4158 
4159 		altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK | M_ZERO);
4160 		error = pf_import_kaltq(pa, altq, IOCPARM_LEN(cmd));
4161 		if (error)
4162 			break;
4163 		altq->local_flags = 0;
4164 
4165 		PF_RULES_WLOCK();
4166 		if (pa->ticket != V_ticket_altqs_inactive) {
4167 			PF_RULES_WUNLOCK();
4168 			free(altq, M_PFALTQ);
4169 			error = EBUSY;
4170 			break;
4171 		}
4172 
4173 		/*
4174 		 * if this is for a queue, find the discipline and
4175 		 * copy the necessary fields
4176 		 */
4177 		if (altq->qname[0] != 0) {
4178 			if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
4179 				PF_RULES_WUNLOCK();
4180 				error = EBUSY;
4181 				free(altq, M_PFALTQ);
4182 				break;
4183 			}
4184 			altq->altq_disc = NULL;
4185 			TAILQ_FOREACH(a, V_pf_altq_ifs_inactive, entries) {
4186 				if (strncmp(a->ifname, altq->ifname,
4187 				    IFNAMSIZ) == 0) {
4188 					altq->altq_disc = a->altq_disc;
4189 					break;
4190 				}
4191 			}
4192 		}
4193 
4194 		if ((ifp = ifunit(altq->ifname)) == NULL)
4195 			altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
4196 		else
4197 			error = altq_add(ifp, altq);
4198 
4199 		if (error) {
4200 			PF_RULES_WUNLOCK();
4201 			free(altq, M_PFALTQ);
4202 			break;
4203 		}
4204 
4205 		if (altq->qname[0] != 0)
4206 			TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
4207 		else
4208 			TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, altq, entries);
4209 		/* version error check done on import above */
4210 		pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
4211 		PF_RULES_WUNLOCK();
4212 		break;
4213 	}
4214 
4215 	case DIOCGETALTQSV0:
4216 	case DIOCGETALTQSV1: {
4217 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
4218 		struct pf_altq		*altq;
4219 
4220 		PF_RULES_RLOCK();
4221 		pa->nr = 0;
4222 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries)
4223 			pa->nr++;
4224 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
4225 			pa->nr++;
4226 		pa->ticket = V_ticket_altqs_active;
4227 		PF_RULES_RUNLOCK();
4228 		break;
4229 	}
4230 
4231 	case DIOCGETALTQV0:
4232 	case DIOCGETALTQV1: {
4233 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
4234 		struct pf_altq		*altq;
4235 
4236 		PF_RULES_RLOCK();
4237 		if (pa->ticket != V_ticket_altqs_active) {
4238 			PF_RULES_RUNLOCK();
4239 			error = EBUSY;
4240 			break;
4241 		}
4242 		altq = pf_altq_get_nth_active(pa->nr);
4243 		if (altq == NULL) {
4244 			PF_RULES_RUNLOCK();
4245 			error = EBUSY;
4246 			break;
4247 		}
4248 		pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
4249 		PF_RULES_RUNLOCK();
4250 		break;
4251 	}
4252 
4253 	case DIOCCHANGEALTQV0:
4254 	case DIOCCHANGEALTQV1:
4255 		/* CHANGEALTQ not supported yet! */
4256 		error = ENODEV;
4257 		break;
4258 
4259 	case DIOCGETQSTATSV0:
4260 	case DIOCGETQSTATSV1: {
4261 		struct pfioc_qstats_v1	*pq = (struct pfioc_qstats_v1 *)addr;
4262 		struct pf_altq		*altq;
4263 		int			 nbytes;
4264 		u_int32_t		 version;
4265 
4266 		PF_RULES_RLOCK();
4267 		if (pq->ticket != V_ticket_altqs_active) {
4268 			PF_RULES_RUNLOCK();
4269 			error = EBUSY;
4270 			break;
4271 		}
4272 		nbytes = pq->nbytes;
4273 		altq = pf_altq_get_nth_active(pq->nr);
4274 		if (altq == NULL) {
4275 			PF_RULES_RUNLOCK();
4276 			error = EBUSY;
4277 			break;
4278 		}
4279 
4280 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
4281 			PF_RULES_RUNLOCK();
4282 			error = ENXIO;
4283 			break;
4284 		}
4285 		PF_RULES_RUNLOCK();
4286 		if (cmd == DIOCGETQSTATSV0)
4287 			version = 0;  /* DIOCGETQSTATSV0 means stats struct v0 */
4288 		else
4289 			version = pq->version;
4290 		error = altq_getqstats(altq, pq->buf, &nbytes, version);
4291 		if (error == 0) {
4292 			pq->scheduler = altq->scheduler;
4293 			pq->nbytes = nbytes;
4294 		}
4295 		break;
4296 	}
4297 #endif /* ALTQ */
4298 
4299 	case DIOCBEGINADDRS: {
4300 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4301 
4302 		error = pf_ioctl_begin_addrs(&pp->ticket);
4303 		break;
4304 	}
4305 
4306 	case DIOCADDADDR: {
4307 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4308 
4309 		error = pf_ioctl_add_addr(pp);
4310 		break;
4311 	}
4312 
4313 	case DIOCGETADDRS: {
4314 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4315 
4316 		error = pf_ioctl_get_addrs(pp);
4317 		break;
4318 	}
4319 
4320 	case DIOCGETADDR: {
4321 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4322 
4323 		error = pf_ioctl_get_addr(pp);
4324 		break;
4325 	}
4326 
4327 	case DIOCCHANGEADDR: {
4328 		struct pfioc_pooladdr	*pca = (struct pfioc_pooladdr *)addr;
4329 		struct pf_kpool		*pool;
4330 		struct pf_kpooladdr	*oldpa = NULL, *newpa = NULL;
4331 		struct pf_kruleset	*ruleset;
4332 		struct pfi_kkif		*kif = NULL;
4333 
4334 		pca->anchor[sizeof(pca->anchor) - 1] = 0;
4335 
4336 		if (pca->action < PF_CHANGE_ADD_HEAD ||
4337 		    pca->action > PF_CHANGE_REMOVE) {
4338 			error = EINVAL;
4339 			break;
4340 		}
4341 		if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
4342 		    pca->addr.addr.type != PF_ADDR_DYNIFTL &&
4343 		    pca->addr.addr.type != PF_ADDR_TABLE) {
4344 			error = EINVAL;
4345 			break;
4346 		}
4347 		if (pca->addr.addr.p.dyn != NULL) {
4348 			error = EINVAL;
4349 			break;
4350 		}
4351 
4352 		if (pca->action != PF_CHANGE_REMOVE) {
4353 #ifndef INET
4354 			if (pca->af == AF_INET) {
4355 				error = EAFNOSUPPORT;
4356 				break;
4357 			}
4358 #endif /* INET */
4359 #ifndef INET6
4360 			if (pca->af == AF_INET6) {
4361 				error = EAFNOSUPPORT;
4362 				break;
4363 			}
4364 #endif /* INET6 */
4365 			newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
4366 			bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
4367 			if (newpa->ifname[0])
4368 				kif = pf_kkif_create(M_WAITOK);
4369 			newpa->kif = NULL;
4370 		}
4371 #define	ERROUT(x)	ERROUT_IOCTL(DIOCCHANGEADDR_error, x)
4372 		PF_RULES_WLOCK();
4373 		ruleset = pf_find_kruleset(pca->anchor);
4374 		if (ruleset == NULL)
4375 			ERROUT(EBUSY);
4376 
4377 		pool = pf_get_kpool(pca->anchor, pca->ticket, pca->r_action,
4378 		    pca->r_num, pca->r_last, 1, 1);
4379 		if (pool == NULL)
4380 			ERROUT(EBUSY);
4381 
4382 		if (pca->action != PF_CHANGE_REMOVE) {
4383 			if (newpa->ifname[0]) {
4384 				newpa->kif = pfi_kkif_attach(kif, newpa->ifname);
4385 				pfi_kkif_ref(newpa->kif);
4386 				kif = NULL;
4387 			}
4388 
4389 			switch (newpa->addr.type) {
4390 			case PF_ADDR_DYNIFTL:
4391 				error = pfi_dynaddr_setup(&newpa->addr,
4392 				    pca->af);
4393 				break;
4394 			case PF_ADDR_TABLE:
4395 				newpa->addr.p.tbl = pfr_attach_table(ruleset,
4396 				    newpa->addr.v.tblname);
4397 				if (newpa->addr.p.tbl == NULL)
4398 					error = ENOMEM;
4399 				break;
4400 			}
4401 			if (error)
4402 				goto DIOCCHANGEADDR_error;
4403 		}
4404 
4405 		switch (pca->action) {
4406 		case PF_CHANGE_ADD_HEAD:
4407 			oldpa = TAILQ_FIRST(&pool->list);
4408 			break;
4409 		case PF_CHANGE_ADD_TAIL:
4410 			oldpa = TAILQ_LAST(&pool->list, pf_kpalist);
4411 			break;
4412 		default:
4413 			oldpa = TAILQ_FIRST(&pool->list);
4414 			for (int i = 0; oldpa && i < pca->nr; i++)
4415 				oldpa = TAILQ_NEXT(oldpa, entries);
4416 
4417 			if (oldpa == NULL)
4418 				ERROUT(EINVAL);
4419 		}
4420 
4421 		if (pca->action == PF_CHANGE_REMOVE) {
4422 			TAILQ_REMOVE(&pool->list, oldpa, entries);
4423 			switch (oldpa->addr.type) {
4424 			case PF_ADDR_DYNIFTL:
4425 				pfi_dynaddr_remove(oldpa->addr.p.dyn);
4426 				break;
4427 			case PF_ADDR_TABLE:
4428 				pfr_detach_table(oldpa->addr.p.tbl);
4429 				break;
4430 			}
4431 			if (oldpa->kif)
4432 				pfi_kkif_unref(oldpa->kif);
4433 			free(oldpa, M_PFRULE);
4434 		} else {
4435 			if (oldpa == NULL)
4436 				TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
4437 			else if (pca->action == PF_CHANGE_ADD_HEAD ||
4438 			    pca->action == PF_CHANGE_ADD_BEFORE)
4439 				TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
4440 			else
4441 				TAILQ_INSERT_AFTER(&pool->list, oldpa,
4442 				    newpa, entries);
4443 		}
4444 
4445 		pool->cur = TAILQ_FIRST(&pool->list);
4446 		PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr, pca->af);
4447 		PF_RULES_WUNLOCK();
4448 		break;
4449 
4450 #undef ERROUT
4451 DIOCCHANGEADDR_error:
4452 		if (newpa != NULL) {
4453 			if (newpa->kif)
4454 				pfi_kkif_unref(newpa->kif);
4455 			free(newpa, M_PFRULE);
4456 		}
4457 		PF_RULES_WUNLOCK();
4458 		pf_kkif_free(kif);
4459 		break;
4460 	}
4461 
4462 	case DIOCGETRULESETS: {
4463 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
4464 		struct pf_kruleset	*ruleset;
4465 		struct pf_kanchor	*anchor;
4466 
4467 		pr->path[sizeof(pr->path) - 1] = 0;
4468 
4469 		PF_RULES_RLOCK();
4470 		if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
4471 			PF_RULES_RUNLOCK();
4472 			error = ENOENT;
4473 			break;
4474 		}
4475 		pr->nr = 0;
4476 		if (ruleset->anchor == NULL) {
4477 			/* XXX kludge for pf_main_ruleset */
4478 			RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
4479 				if (anchor->parent == NULL)
4480 					pr->nr++;
4481 		} else {
4482 			RB_FOREACH(anchor, pf_kanchor_node,
4483 			    &ruleset->anchor->children)
4484 				pr->nr++;
4485 		}
4486 		PF_RULES_RUNLOCK();
4487 		break;
4488 	}
4489 
4490 	case DIOCGETRULESET: {
4491 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
4492 		struct pf_kruleset	*ruleset;
4493 		struct pf_kanchor	*anchor;
4494 		u_int32_t		 nr = 0;
4495 
4496 		pr->path[sizeof(pr->path) - 1] = 0;
4497 
4498 		PF_RULES_RLOCK();
4499 		if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
4500 			PF_RULES_RUNLOCK();
4501 			error = ENOENT;
4502 			break;
4503 		}
4504 		pr->name[0] = 0;
4505 		if (ruleset->anchor == NULL) {
4506 			/* XXX kludge for pf_main_ruleset */
4507 			RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
4508 				if (anchor->parent == NULL && nr++ == pr->nr) {
4509 					strlcpy(pr->name, anchor->name,
4510 					    sizeof(pr->name));
4511 					break;
4512 				}
4513 		} else {
4514 			RB_FOREACH(anchor, pf_kanchor_node,
4515 			    &ruleset->anchor->children)
4516 				if (nr++ == pr->nr) {
4517 					strlcpy(pr->name, anchor->name,
4518 					    sizeof(pr->name));
4519 					break;
4520 				}
4521 		}
4522 		if (!pr->name[0])
4523 			error = EBUSY;
4524 		PF_RULES_RUNLOCK();
4525 		break;
4526 	}
4527 
4528 	case DIOCRCLRTABLES: {
4529 		struct pfioc_table *io = (struct pfioc_table *)addr;
4530 
4531 		if (io->pfrio_esize != 0) {
4532 			error = ENODEV;
4533 			break;
4534 		}
4535 		PF_RULES_WLOCK();
4536 		error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
4537 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
4538 		PF_RULES_WUNLOCK();
4539 		break;
4540 	}
4541 
4542 	case DIOCRADDTABLES: {
4543 		struct pfioc_table *io = (struct pfioc_table *)addr;
4544 		struct pfr_table *pfrts;
4545 		size_t totlen;
4546 
4547 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4548 			error = ENODEV;
4549 			break;
4550 		}
4551 
4552 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4553 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4554 			error = ENOMEM;
4555 			break;
4556 		}
4557 
4558 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4559 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4560 		    M_TEMP, M_WAITOK);
4561 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4562 		if (error) {
4563 			free(pfrts, M_TEMP);
4564 			break;
4565 		}
4566 		PF_RULES_WLOCK();
4567 		error = pfr_add_tables(pfrts, io->pfrio_size,
4568 		    &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4569 		PF_RULES_WUNLOCK();
4570 		free(pfrts, M_TEMP);
4571 		break;
4572 	}
4573 
4574 	case DIOCRDELTABLES: {
4575 		struct pfioc_table *io = (struct pfioc_table *)addr;
4576 		struct pfr_table *pfrts;
4577 		size_t totlen;
4578 
4579 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4580 			error = ENODEV;
4581 			break;
4582 		}
4583 
4584 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4585 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4586 			error = ENOMEM;
4587 			break;
4588 		}
4589 
4590 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4591 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4592 		    M_TEMP, M_WAITOK);
4593 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4594 		if (error) {
4595 			free(pfrts, M_TEMP);
4596 			break;
4597 		}
4598 		PF_RULES_WLOCK();
4599 		error = pfr_del_tables(pfrts, io->pfrio_size,
4600 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4601 		PF_RULES_WUNLOCK();
4602 		free(pfrts, M_TEMP);
4603 		break;
4604 	}
4605 
4606 	case DIOCRGETTABLES: {
4607 		struct pfioc_table *io = (struct pfioc_table *)addr;
4608 		struct pfr_table *pfrts;
4609 		size_t totlen;
4610 		int n;
4611 
4612 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4613 			error = ENODEV;
4614 			break;
4615 		}
4616 		PF_RULES_RLOCK();
4617 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4618 		if (n < 0) {
4619 			PF_RULES_RUNLOCK();
4620 			error = EINVAL;
4621 			break;
4622 		}
4623 		io->pfrio_size = min(io->pfrio_size, n);
4624 
4625 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4626 
4627 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4628 		    M_TEMP, M_NOWAIT | M_ZERO);
4629 		if (pfrts == NULL) {
4630 			error = ENOMEM;
4631 			PF_RULES_RUNLOCK();
4632 			break;
4633 		}
4634 		error = pfr_get_tables(&io->pfrio_table, pfrts,
4635 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4636 		PF_RULES_RUNLOCK();
4637 		if (error == 0)
4638 			error = copyout(pfrts, io->pfrio_buffer, totlen);
4639 		free(pfrts, M_TEMP);
4640 		break;
4641 	}
4642 
4643 	case DIOCRGETTSTATS: {
4644 		struct pfioc_table *io = (struct pfioc_table *)addr;
4645 		struct pfr_tstats *pfrtstats;
4646 		size_t totlen;
4647 		int n;
4648 
4649 		if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
4650 			error = ENODEV;
4651 			break;
4652 		}
4653 		PF_TABLE_STATS_LOCK();
4654 		PF_RULES_RLOCK();
4655 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4656 		if (n < 0) {
4657 			PF_RULES_RUNLOCK();
4658 			PF_TABLE_STATS_UNLOCK();
4659 			error = EINVAL;
4660 			break;
4661 		}
4662 		io->pfrio_size = min(io->pfrio_size, n);
4663 
4664 		totlen = io->pfrio_size * sizeof(struct pfr_tstats);
4665 		pfrtstats = mallocarray(io->pfrio_size,
4666 		    sizeof(struct pfr_tstats), M_TEMP, M_NOWAIT | M_ZERO);
4667 		if (pfrtstats == NULL) {
4668 			error = ENOMEM;
4669 			PF_RULES_RUNLOCK();
4670 			PF_TABLE_STATS_UNLOCK();
4671 			break;
4672 		}
4673 		error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
4674 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4675 		PF_RULES_RUNLOCK();
4676 		PF_TABLE_STATS_UNLOCK();
4677 		if (error == 0)
4678 			error = copyout(pfrtstats, io->pfrio_buffer, totlen);
4679 		free(pfrtstats, M_TEMP);
4680 		break;
4681 	}
4682 
4683 	case DIOCRCLRTSTATS: {
4684 		struct pfioc_table *io = (struct pfioc_table *)addr;
4685 		struct pfr_table *pfrts;
4686 		size_t totlen;
4687 
4688 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4689 			error = ENODEV;
4690 			break;
4691 		}
4692 
4693 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4694 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4695 			/* We used to count tables and use the minimum required
4696 			 * size, so we didn't fail on overly large requests.
4697 			 * Keep doing so. */
4698 			io->pfrio_size = pf_ioctl_maxcount;
4699 			break;
4700 		}
4701 
4702 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4703 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4704 		    M_TEMP, M_WAITOK);
4705 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4706 		if (error) {
4707 			free(pfrts, M_TEMP);
4708 			break;
4709 		}
4710 
4711 		PF_TABLE_STATS_LOCK();
4712 		PF_RULES_RLOCK();
4713 		error = pfr_clr_tstats(pfrts, io->pfrio_size,
4714 		    &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4715 		PF_RULES_RUNLOCK();
4716 		PF_TABLE_STATS_UNLOCK();
4717 		free(pfrts, M_TEMP);
4718 		break;
4719 	}
4720 
4721 	case DIOCRSETTFLAGS: {
4722 		struct pfioc_table *io = (struct pfioc_table *)addr;
4723 		struct pfr_table *pfrts;
4724 		size_t totlen;
4725 		int n;
4726 
4727 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4728 			error = ENODEV;
4729 			break;
4730 		}
4731 
4732 		PF_RULES_RLOCK();
4733 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4734 		if (n < 0) {
4735 			PF_RULES_RUNLOCK();
4736 			error = EINVAL;
4737 			break;
4738 		}
4739 
4740 		io->pfrio_size = min(io->pfrio_size, n);
4741 		PF_RULES_RUNLOCK();
4742 
4743 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4744 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4745 		    M_TEMP, M_WAITOK);
4746 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4747 		if (error) {
4748 			free(pfrts, M_TEMP);
4749 			break;
4750 		}
4751 		PF_RULES_WLOCK();
4752 		error = pfr_set_tflags(pfrts, io->pfrio_size,
4753 		    io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
4754 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4755 		PF_RULES_WUNLOCK();
4756 		free(pfrts, M_TEMP);
4757 		break;
4758 	}
4759 
4760 	case DIOCRCLRADDRS: {
4761 		struct pfioc_table *io = (struct pfioc_table *)addr;
4762 
4763 		if (io->pfrio_esize != 0) {
4764 			error = ENODEV;
4765 			break;
4766 		}
4767 		PF_RULES_WLOCK();
4768 		error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
4769 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
4770 		PF_RULES_WUNLOCK();
4771 		break;
4772 	}
4773 
4774 	case DIOCRADDADDRS: {
4775 		struct pfioc_table *io = (struct pfioc_table *)addr;
4776 		struct pfr_addr *pfras;
4777 		size_t totlen;
4778 
4779 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4780 			error = ENODEV;
4781 			break;
4782 		}
4783 		if (io->pfrio_size < 0 ||
4784 		    io->pfrio_size > pf_ioctl_maxcount ||
4785 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4786 			error = EINVAL;
4787 			break;
4788 		}
4789 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4790 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4791 		    M_TEMP, M_WAITOK);
4792 		error = copyin(io->pfrio_buffer, pfras, totlen);
4793 		if (error) {
4794 			free(pfras, M_TEMP);
4795 			break;
4796 		}
4797 		PF_RULES_WLOCK();
4798 		error = pfr_add_addrs(&io->pfrio_table, pfras,
4799 		    io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
4800 		    PFR_FLAG_USERIOCTL);
4801 		PF_RULES_WUNLOCK();
4802 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4803 			error = copyout(pfras, io->pfrio_buffer, totlen);
4804 		free(pfras, M_TEMP);
4805 		break;
4806 	}
4807 
4808 	case DIOCRDELADDRS: {
4809 		struct pfioc_table *io = (struct pfioc_table *)addr;
4810 		struct pfr_addr *pfras;
4811 		size_t totlen;
4812 
4813 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4814 			error = ENODEV;
4815 			break;
4816 		}
4817 		if (io->pfrio_size < 0 ||
4818 		    io->pfrio_size > pf_ioctl_maxcount ||
4819 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4820 			error = EINVAL;
4821 			break;
4822 		}
4823 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4824 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4825 		    M_TEMP, M_WAITOK);
4826 		error = copyin(io->pfrio_buffer, pfras, totlen);
4827 		if (error) {
4828 			free(pfras, M_TEMP);
4829 			break;
4830 		}
4831 		PF_RULES_WLOCK();
4832 		error = pfr_del_addrs(&io->pfrio_table, pfras,
4833 		    io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
4834 		    PFR_FLAG_USERIOCTL);
4835 		PF_RULES_WUNLOCK();
4836 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4837 			error = copyout(pfras, io->pfrio_buffer, totlen);
4838 		free(pfras, M_TEMP);
4839 		break;
4840 	}
4841 
4842 	case DIOCRSETADDRS: {
4843 		struct pfioc_table *io = (struct pfioc_table *)addr;
4844 		struct pfr_addr *pfras;
4845 		size_t totlen, count;
4846 
4847 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4848 			error = ENODEV;
4849 			break;
4850 		}
4851 		if (io->pfrio_size < 0 || io->pfrio_size2 < 0) {
4852 			error = EINVAL;
4853 			break;
4854 		}
4855 		count = max(io->pfrio_size, io->pfrio_size2);
4856 		if (count > pf_ioctl_maxcount ||
4857 		    WOULD_OVERFLOW(count, sizeof(struct pfr_addr))) {
4858 			error = EINVAL;
4859 			break;
4860 		}
4861 		totlen = count * sizeof(struct pfr_addr);
4862 		pfras = mallocarray(count, sizeof(struct pfr_addr), M_TEMP,
4863 		    M_WAITOK);
4864 		error = copyin(io->pfrio_buffer, pfras, totlen);
4865 		if (error) {
4866 			free(pfras, M_TEMP);
4867 			break;
4868 		}
4869 		PF_RULES_WLOCK();
4870 		error = pfr_set_addrs(&io->pfrio_table, pfras,
4871 		    io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
4872 		    &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
4873 		    PFR_FLAG_USERIOCTL, 0);
4874 		PF_RULES_WUNLOCK();
4875 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4876 			error = copyout(pfras, io->pfrio_buffer, totlen);
4877 		free(pfras, M_TEMP);
4878 		break;
4879 	}
4880 
4881 	case DIOCRGETADDRS: {
4882 		struct pfioc_table *io = (struct pfioc_table *)addr;
4883 		struct pfr_addr *pfras;
4884 		size_t totlen;
4885 
4886 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4887 			error = ENODEV;
4888 			break;
4889 		}
4890 		if (io->pfrio_size < 0 ||
4891 		    io->pfrio_size > pf_ioctl_maxcount ||
4892 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4893 			error = EINVAL;
4894 			break;
4895 		}
4896 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4897 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4898 		    M_TEMP, M_WAITOK | M_ZERO);
4899 		PF_RULES_RLOCK();
4900 		error = pfr_get_addrs(&io->pfrio_table, pfras,
4901 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4902 		PF_RULES_RUNLOCK();
4903 		if (error == 0)
4904 			error = copyout(pfras, io->pfrio_buffer, totlen);
4905 		free(pfras, M_TEMP);
4906 		break;
4907 	}
4908 
4909 	case DIOCRGETASTATS: {
4910 		struct pfioc_table *io = (struct pfioc_table *)addr;
4911 		struct pfr_astats *pfrastats;
4912 		size_t totlen;
4913 
4914 		if (io->pfrio_esize != sizeof(struct pfr_astats)) {
4915 			error = ENODEV;
4916 			break;
4917 		}
4918 		if (io->pfrio_size < 0 ||
4919 		    io->pfrio_size > pf_ioctl_maxcount ||
4920 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_astats))) {
4921 			error = EINVAL;
4922 			break;
4923 		}
4924 		totlen = io->pfrio_size * sizeof(struct pfr_astats);
4925 		pfrastats = mallocarray(io->pfrio_size,
4926 		    sizeof(struct pfr_astats), M_TEMP, M_WAITOK | M_ZERO);
4927 		PF_RULES_RLOCK();
4928 		error = pfr_get_astats(&io->pfrio_table, pfrastats,
4929 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4930 		PF_RULES_RUNLOCK();
4931 		if (error == 0)
4932 			error = copyout(pfrastats, io->pfrio_buffer, totlen);
4933 		free(pfrastats, M_TEMP);
4934 		break;
4935 	}
4936 
4937 	case DIOCRCLRASTATS: {
4938 		struct pfioc_table *io = (struct pfioc_table *)addr;
4939 		struct pfr_addr *pfras;
4940 		size_t totlen;
4941 
4942 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4943 			error = ENODEV;
4944 			break;
4945 		}
4946 		if (io->pfrio_size < 0 ||
4947 		    io->pfrio_size > pf_ioctl_maxcount ||
4948 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4949 			error = EINVAL;
4950 			break;
4951 		}
4952 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4953 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4954 		    M_TEMP, M_WAITOK);
4955 		error = copyin(io->pfrio_buffer, pfras, totlen);
4956 		if (error) {
4957 			free(pfras, M_TEMP);
4958 			break;
4959 		}
4960 		PF_RULES_WLOCK();
4961 		error = pfr_clr_astats(&io->pfrio_table, pfras,
4962 		    io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
4963 		    PFR_FLAG_USERIOCTL);
4964 		PF_RULES_WUNLOCK();
4965 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4966 			error = copyout(pfras, io->pfrio_buffer, totlen);
4967 		free(pfras, M_TEMP);
4968 		break;
4969 	}
4970 
4971 	case DIOCRTSTADDRS: {
4972 		struct pfioc_table *io = (struct pfioc_table *)addr;
4973 		struct pfr_addr *pfras;
4974 		size_t totlen;
4975 
4976 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4977 			error = ENODEV;
4978 			break;
4979 		}
4980 		if (io->pfrio_size < 0 ||
4981 		    io->pfrio_size > pf_ioctl_maxcount ||
4982 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4983 			error = EINVAL;
4984 			break;
4985 		}
4986 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4987 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4988 		    M_TEMP, M_WAITOK);
4989 		error = copyin(io->pfrio_buffer, pfras, totlen);
4990 		if (error) {
4991 			free(pfras, M_TEMP);
4992 			break;
4993 		}
4994 		PF_RULES_RLOCK();
4995 		error = pfr_tst_addrs(&io->pfrio_table, pfras,
4996 		    io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
4997 		    PFR_FLAG_USERIOCTL);
4998 		PF_RULES_RUNLOCK();
4999 		if (error == 0)
5000 			error = copyout(pfras, io->pfrio_buffer, totlen);
5001 		free(pfras, M_TEMP);
5002 		break;
5003 	}
5004 
5005 	case DIOCRINADEFINE: {
5006 		struct pfioc_table *io = (struct pfioc_table *)addr;
5007 		struct pfr_addr *pfras;
5008 		size_t totlen;
5009 
5010 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5011 			error = ENODEV;
5012 			break;
5013 		}
5014 		if (io->pfrio_size < 0 ||
5015 		    io->pfrio_size > pf_ioctl_maxcount ||
5016 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
5017 			error = EINVAL;
5018 			break;
5019 		}
5020 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
5021 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5022 		    M_TEMP, M_WAITOK);
5023 		error = copyin(io->pfrio_buffer, pfras, totlen);
5024 		if (error) {
5025 			free(pfras, M_TEMP);
5026 			break;
5027 		}
5028 		PF_RULES_WLOCK();
5029 		error = pfr_ina_define(&io->pfrio_table, pfras,
5030 		    io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
5031 		    io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
5032 		PF_RULES_WUNLOCK();
5033 		free(pfras, M_TEMP);
5034 		break;
5035 	}
5036 
5037 	case DIOCOSFPADD: {
5038 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
5039 		PF_RULES_WLOCK();
5040 		error = pf_osfp_add(io);
5041 		PF_RULES_WUNLOCK();
5042 		break;
5043 	}
5044 
5045 	case DIOCOSFPGET: {
5046 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
5047 		PF_RULES_RLOCK();
5048 		error = pf_osfp_get(io);
5049 		PF_RULES_RUNLOCK();
5050 		break;
5051 	}
5052 
5053 	case DIOCXBEGIN: {
5054 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
5055 		struct pfioc_trans_e	*ioes, *ioe;
5056 		size_t			 totlen;
5057 		int			 i;
5058 
5059 		if (io->esize != sizeof(*ioe)) {
5060 			error = ENODEV;
5061 			break;
5062 		}
5063 		if (io->size < 0 ||
5064 		    io->size > pf_ioctl_maxcount ||
5065 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5066 			error = EINVAL;
5067 			break;
5068 		}
5069 		totlen = sizeof(struct pfioc_trans_e) * io->size;
5070 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5071 		    M_TEMP, M_WAITOK);
5072 		error = copyin(io->array, ioes, totlen);
5073 		if (error) {
5074 			free(ioes, M_TEMP);
5075 			break;
5076 		}
5077 		/* Ensure there's no more ethernet rules to clean up. */
5078 		NET_EPOCH_DRAIN_CALLBACKS();
5079 		PF_RULES_WLOCK();
5080 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5081 			ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
5082 			switch (ioe->rs_num) {
5083 			case PF_RULESET_ETH:
5084 				if ((error = pf_begin_eth(&ioe->ticket, ioe->anchor))) {
5085 					PF_RULES_WUNLOCK();
5086 					free(ioes, M_TEMP);
5087 					goto fail;
5088 				}
5089 				break;
5090 #ifdef ALTQ
5091 			case PF_RULESET_ALTQ:
5092 				if (ioe->anchor[0]) {
5093 					PF_RULES_WUNLOCK();
5094 					free(ioes, M_TEMP);
5095 					error = EINVAL;
5096 					goto fail;
5097 				}
5098 				if ((error = pf_begin_altq(&ioe->ticket))) {
5099 					PF_RULES_WUNLOCK();
5100 					free(ioes, M_TEMP);
5101 					goto fail;
5102 				}
5103 				break;
5104 #endif /* ALTQ */
5105 			case PF_RULESET_TABLE:
5106 			    {
5107 				struct pfr_table table;
5108 
5109 				bzero(&table, sizeof(table));
5110 				strlcpy(table.pfrt_anchor, ioe->anchor,
5111 				    sizeof(table.pfrt_anchor));
5112 				if ((error = pfr_ina_begin(&table,
5113 				    &ioe->ticket, NULL, 0))) {
5114 					PF_RULES_WUNLOCK();
5115 					free(ioes, M_TEMP);
5116 					goto fail;
5117 				}
5118 				break;
5119 			    }
5120 			default:
5121 				if ((error = pf_begin_rules(&ioe->ticket,
5122 				    ioe->rs_num, ioe->anchor))) {
5123 					PF_RULES_WUNLOCK();
5124 					free(ioes, M_TEMP);
5125 					goto fail;
5126 				}
5127 				break;
5128 			}
5129 		}
5130 		PF_RULES_WUNLOCK();
5131 		error = copyout(ioes, io->array, totlen);
5132 		free(ioes, M_TEMP);
5133 		break;
5134 	}
5135 
5136 	case DIOCXROLLBACK: {
5137 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
5138 		struct pfioc_trans_e	*ioe, *ioes;
5139 		size_t			 totlen;
5140 		int			 i;
5141 
5142 		if (io->esize != sizeof(*ioe)) {
5143 			error = ENODEV;
5144 			break;
5145 		}
5146 		if (io->size < 0 ||
5147 		    io->size > pf_ioctl_maxcount ||
5148 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5149 			error = EINVAL;
5150 			break;
5151 		}
5152 		totlen = sizeof(struct pfioc_trans_e) * io->size;
5153 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5154 		    M_TEMP, M_WAITOK);
5155 		error = copyin(io->array, ioes, totlen);
5156 		if (error) {
5157 			free(ioes, M_TEMP);
5158 			break;
5159 		}
5160 		PF_RULES_WLOCK();
5161 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5162 			ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
5163 			switch (ioe->rs_num) {
5164 			case PF_RULESET_ETH:
5165 				if ((error = pf_rollback_eth(ioe->ticket,
5166 				    ioe->anchor))) {
5167 					PF_RULES_WUNLOCK();
5168 					free(ioes, M_TEMP);
5169 					goto fail; /* really bad */
5170 				}
5171 				break;
5172 #ifdef ALTQ
5173 			case PF_RULESET_ALTQ:
5174 				if (ioe->anchor[0]) {
5175 					PF_RULES_WUNLOCK();
5176 					free(ioes, M_TEMP);
5177 					error = EINVAL;
5178 					goto fail;
5179 				}
5180 				if ((error = pf_rollback_altq(ioe->ticket))) {
5181 					PF_RULES_WUNLOCK();
5182 					free(ioes, M_TEMP);
5183 					goto fail; /* really bad */
5184 				}
5185 				break;
5186 #endif /* ALTQ */
5187 			case PF_RULESET_TABLE:
5188 			    {
5189 				struct pfr_table table;
5190 
5191 				bzero(&table, sizeof(table));
5192 				strlcpy(table.pfrt_anchor, ioe->anchor,
5193 				    sizeof(table.pfrt_anchor));
5194 				if ((error = pfr_ina_rollback(&table,
5195 				    ioe->ticket, NULL, 0))) {
5196 					PF_RULES_WUNLOCK();
5197 					free(ioes, M_TEMP);
5198 					goto fail; /* really bad */
5199 				}
5200 				break;
5201 			    }
5202 			default:
5203 				if ((error = pf_rollback_rules(ioe->ticket,
5204 				    ioe->rs_num, ioe->anchor))) {
5205 					PF_RULES_WUNLOCK();
5206 					free(ioes, M_TEMP);
5207 					goto fail; /* really bad */
5208 				}
5209 				break;
5210 			}
5211 		}
5212 		PF_RULES_WUNLOCK();
5213 		free(ioes, M_TEMP);
5214 		break;
5215 	}
5216 
5217 	case DIOCXCOMMIT: {
5218 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
5219 		struct pfioc_trans_e	*ioe, *ioes;
5220 		struct pf_kruleset	*rs;
5221 		struct pf_keth_ruleset	*ers;
5222 		size_t			 totlen;
5223 		int			 i;
5224 
5225 		if (io->esize != sizeof(*ioe)) {
5226 			error = ENODEV;
5227 			break;
5228 		}
5229 
5230 		if (io->size < 0 ||
5231 		    io->size > pf_ioctl_maxcount ||
5232 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5233 			error = EINVAL;
5234 			break;
5235 		}
5236 
5237 		totlen = sizeof(struct pfioc_trans_e) * io->size;
5238 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5239 		    M_TEMP, M_WAITOK);
5240 		error = copyin(io->array, ioes, totlen);
5241 		if (error) {
5242 			free(ioes, M_TEMP);
5243 			break;
5244 		}
5245 		PF_RULES_WLOCK();
5246 		/* First makes sure everything will succeed. */
5247 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5248 			ioe->anchor[sizeof(ioe->anchor) - 1] = 0;
5249 			switch (ioe->rs_num) {
5250 			case PF_RULESET_ETH:
5251 				ers = pf_find_keth_ruleset(ioe->anchor);
5252 				if (ers == NULL || ioe->ticket == 0 ||
5253 				    ioe->ticket != ers->inactive.ticket) {
5254 					PF_RULES_WUNLOCK();
5255 					free(ioes, M_TEMP);
5256 					error = EINVAL;
5257 					goto fail;
5258 				}
5259 				break;
5260 #ifdef ALTQ
5261 			case PF_RULESET_ALTQ:
5262 				if (ioe->anchor[0]) {
5263 					PF_RULES_WUNLOCK();
5264 					free(ioes, M_TEMP);
5265 					error = EINVAL;
5266 					goto fail;
5267 				}
5268 				if (!V_altqs_inactive_open || ioe->ticket !=
5269 				    V_ticket_altqs_inactive) {
5270 					PF_RULES_WUNLOCK();
5271 					free(ioes, M_TEMP);
5272 					error = EBUSY;
5273 					goto fail;
5274 				}
5275 				break;
5276 #endif /* ALTQ */
5277 			case PF_RULESET_TABLE:
5278 				rs = pf_find_kruleset(ioe->anchor);
5279 				if (rs == NULL || !rs->topen || ioe->ticket !=
5280 				    rs->tticket) {
5281 					PF_RULES_WUNLOCK();
5282 					free(ioes, M_TEMP);
5283 					error = EBUSY;
5284 					goto fail;
5285 				}
5286 				break;
5287 			default:
5288 				if (ioe->rs_num < 0 || ioe->rs_num >=
5289 				    PF_RULESET_MAX) {
5290 					PF_RULES_WUNLOCK();
5291 					free(ioes, M_TEMP);
5292 					error = EINVAL;
5293 					goto fail;
5294 				}
5295 				rs = pf_find_kruleset(ioe->anchor);
5296 				if (rs == NULL ||
5297 				    !rs->rules[ioe->rs_num].inactive.open ||
5298 				    rs->rules[ioe->rs_num].inactive.ticket !=
5299 				    ioe->ticket) {
5300 					PF_RULES_WUNLOCK();
5301 					free(ioes, M_TEMP);
5302 					error = EBUSY;
5303 					goto fail;
5304 				}
5305 				break;
5306 			}
5307 		}
5308 		/* Now do the commit - no errors should happen here. */
5309 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5310 			switch (ioe->rs_num) {
5311 			case PF_RULESET_ETH:
5312 				if ((error = pf_commit_eth(ioe->ticket, ioe->anchor))) {
5313 					PF_RULES_WUNLOCK();
5314 					free(ioes, M_TEMP);
5315 					goto fail; /* really bad */
5316 				}
5317 				break;
5318 #ifdef ALTQ
5319 			case PF_RULESET_ALTQ:
5320 				if ((error = pf_commit_altq(ioe->ticket))) {
5321 					PF_RULES_WUNLOCK();
5322 					free(ioes, M_TEMP);
5323 					goto fail; /* really bad */
5324 				}
5325 				break;
5326 #endif /* ALTQ */
5327 			case PF_RULESET_TABLE:
5328 			    {
5329 				struct pfr_table table;
5330 
5331 				bzero(&table, sizeof(table));
5332 				(void)strlcpy(table.pfrt_anchor, ioe->anchor,
5333 				    sizeof(table.pfrt_anchor));
5334 				if ((error = pfr_ina_commit(&table,
5335 				    ioe->ticket, NULL, NULL, 0))) {
5336 					PF_RULES_WUNLOCK();
5337 					free(ioes, M_TEMP);
5338 					goto fail; /* really bad */
5339 				}
5340 				break;
5341 			    }
5342 			default:
5343 				if ((error = pf_commit_rules(ioe->ticket,
5344 				    ioe->rs_num, ioe->anchor))) {
5345 					PF_RULES_WUNLOCK();
5346 					free(ioes, M_TEMP);
5347 					goto fail; /* really bad */
5348 				}
5349 				break;
5350 			}
5351 		}
5352 		PF_RULES_WUNLOCK();
5353 
5354 		/* Only hook into EtherNet taffic if we've got rules for it. */
5355 		if (! TAILQ_EMPTY(V_pf_keth->active.rules))
5356 			hook_pf_eth();
5357 		else
5358 			dehook_pf_eth();
5359 
5360 		free(ioes, M_TEMP);
5361 		break;
5362 	}
5363 
5364 	case DIOCGETSRCNODES: {
5365 		struct pfioc_src_nodes	*psn = (struct pfioc_src_nodes *)addr;
5366 		struct pf_srchash	*sh;
5367 		struct pf_ksrc_node	*n;
5368 		struct pf_src_node	*p, *pstore;
5369 		uint32_t		 i, nr = 0;
5370 
5371 		for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask;
5372 				i++, sh++) {
5373 			PF_HASHROW_LOCK(sh);
5374 			LIST_FOREACH(n, &sh->nodes, entry)
5375 				nr++;
5376 			PF_HASHROW_UNLOCK(sh);
5377 		}
5378 
5379 		psn->psn_len = min(psn->psn_len,
5380 		    sizeof(struct pf_src_node) * nr);
5381 
5382 		if (psn->psn_len == 0) {
5383 			psn->psn_len = sizeof(struct pf_src_node) * nr;
5384 			break;
5385 		}
5386 
5387 		nr = 0;
5388 
5389 		p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK | M_ZERO);
5390 		for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask;
5391 		    i++, sh++) {
5392 		    PF_HASHROW_LOCK(sh);
5393 		    LIST_FOREACH(n, &sh->nodes, entry) {
5394 
5395 			if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
5396 				break;
5397 
5398 			pf_src_node_copy(n, p);
5399 
5400 			p++;
5401 			nr++;
5402 		    }
5403 		    PF_HASHROW_UNLOCK(sh);
5404 		}
5405 		error = copyout(pstore, psn->psn_src_nodes,
5406 		    sizeof(struct pf_src_node) * nr);
5407 		if (error) {
5408 			free(pstore, M_TEMP);
5409 			break;
5410 		}
5411 		psn->psn_len = sizeof(struct pf_src_node) * nr;
5412 		free(pstore, M_TEMP);
5413 		break;
5414 	}
5415 
5416 	case DIOCCLRSRCNODES: {
5417 		pf_clear_srcnodes(NULL);
5418 		pf_purge_expired_src_nodes();
5419 		break;
5420 	}
5421 
5422 	case DIOCKILLSRCNODES:
5423 		pf_kill_srcnodes((struct pfioc_src_node_kill *)addr);
5424 		break;
5425 
5426 #ifdef COMPAT_FREEBSD13
5427 	case DIOCKEEPCOUNTERS_FREEBSD13:
5428 #endif
5429 	case DIOCKEEPCOUNTERS:
5430 		error = pf_keepcounters((struct pfioc_nv *)addr);
5431 		break;
5432 
5433 	case DIOCGETSYNCOOKIES:
5434 		error = pf_get_syncookies((struct pfioc_nv *)addr);
5435 		break;
5436 
5437 	case DIOCSETSYNCOOKIES:
5438 		error = pf_set_syncookies((struct pfioc_nv *)addr);
5439 		break;
5440 
5441 	case DIOCSETHOSTID: {
5442 		u_int32_t	*hostid = (u_int32_t *)addr;
5443 
5444 		PF_RULES_WLOCK();
5445 		if (*hostid == 0)
5446 			V_pf_status.hostid = arc4random();
5447 		else
5448 			V_pf_status.hostid = *hostid;
5449 		PF_RULES_WUNLOCK();
5450 		break;
5451 	}
5452 
5453 	case DIOCOSFPFLUSH:
5454 		PF_RULES_WLOCK();
5455 		pf_osfp_flush();
5456 		PF_RULES_WUNLOCK();
5457 		break;
5458 
5459 	case DIOCIGETIFACES: {
5460 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
5461 		struct pfi_kif *ifstore;
5462 		size_t bufsiz;
5463 
5464 		if (io->pfiio_esize != sizeof(struct pfi_kif)) {
5465 			error = ENODEV;
5466 			break;
5467 		}
5468 
5469 		if (io->pfiio_size < 0 ||
5470 		    io->pfiio_size > pf_ioctl_maxcount ||
5471 		    WOULD_OVERFLOW(io->pfiio_size, sizeof(struct pfi_kif))) {
5472 			error = EINVAL;
5473 			break;
5474 		}
5475 
5476 		io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5477 
5478 		bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
5479 		ifstore = mallocarray(io->pfiio_size, sizeof(struct pfi_kif),
5480 		    M_TEMP, M_WAITOK | M_ZERO);
5481 
5482 		PF_RULES_RLOCK();
5483 		pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
5484 		PF_RULES_RUNLOCK();
5485 		error = copyout(ifstore, io->pfiio_buffer, bufsiz);
5486 		free(ifstore, M_TEMP);
5487 		break;
5488 	}
5489 
5490 	case DIOCSETIFFLAG: {
5491 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
5492 
5493 		io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5494 
5495 		PF_RULES_WLOCK();
5496 		error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
5497 		PF_RULES_WUNLOCK();
5498 		break;
5499 	}
5500 
5501 	case DIOCCLRIFFLAG: {
5502 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
5503 
5504 		io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5505 
5506 		PF_RULES_WLOCK();
5507 		error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
5508 		PF_RULES_WUNLOCK();
5509 		break;
5510 	}
5511 
5512 	case DIOCSETREASS: {
5513 		u_int32_t	*reass = (u_int32_t *)addr;
5514 
5515 		V_pf_status.reass = *reass & (PF_REASS_ENABLED|PF_REASS_NODF);
5516 		/* Removal of DF flag without reassembly enabled is not a
5517 		 * valid combination. Disable reassembly in such case. */
5518 		if (!(V_pf_status.reass & PF_REASS_ENABLED))
5519 			V_pf_status.reass = 0;
5520 		break;
5521 	}
5522 
5523 	default:
5524 		error = ENODEV;
5525 		break;
5526 	}
5527 fail:
5528 	CURVNET_RESTORE();
5529 
5530 #undef ERROUT_IOCTL
5531 
5532 	return (error);
5533 }
5534 
5535 void
5536 pfsync_state_export(union pfsync_state_union *sp, struct pf_kstate *st, int msg_version)
5537 {
5538 	bzero(sp, sizeof(union pfsync_state_union));
5539 
5540 	/* copy from state key */
5541 	sp->pfs_1301.key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
5542 	sp->pfs_1301.key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
5543 	sp->pfs_1301.key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
5544 	sp->pfs_1301.key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
5545 	sp->pfs_1301.key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
5546 	sp->pfs_1301.key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
5547 	sp->pfs_1301.key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
5548 	sp->pfs_1301.key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
5549 	sp->pfs_1301.proto = st->key[PF_SK_WIRE]->proto;
5550 	sp->pfs_1301.af = st->key[PF_SK_WIRE]->af;
5551 
5552 	/* copy from state */
5553 	strlcpy(sp->pfs_1301.ifname, st->kif->pfik_name, sizeof(sp->pfs_1301.ifname));
5554 	bcopy(&st->rt_addr, &sp->pfs_1301.rt_addr, sizeof(sp->pfs_1301.rt_addr));
5555 	sp->pfs_1301.creation = htonl(time_uptime - (st->creation / 1000));
5556 	sp->pfs_1301.expire = pf_state_expires(st);
5557 	if (sp->pfs_1301.expire <= time_uptime)
5558 		sp->pfs_1301.expire = htonl(0);
5559 	else
5560 		sp->pfs_1301.expire = htonl(sp->pfs_1301.expire - time_uptime);
5561 
5562 	sp->pfs_1301.direction = st->direction;
5563 	sp->pfs_1301.log = st->act.log;
5564 	sp->pfs_1301.timeout = st->timeout;
5565 
5566 	switch (msg_version) {
5567 		case PFSYNC_MSG_VERSION_1301:
5568 			sp->pfs_1301.state_flags = st->state_flags;
5569 			break;
5570 		case PFSYNC_MSG_VERSION_1400:
5571 			sp->pfs_1400.state_flags = htons(st->state_flags);
5572 			sp->pfs_1400.qid = htons(st->act.qid);
5573 			sp->pfs_1400.pqid = htons(st->act.pqid);
5574 			sp->pfs_1400.dnpipe = htons(st->act.dnpipe);
5575 			sp->pfs_1400.dnrpipe = htons(st->act.dnrpipe);
5576 			sp->pfs_1400.rtableid = htonl(st->act.rtableid);
5577 			sp->pfs_1400.min_ttl = st->act.min_ttl;
5578 			sp->pfs_1400.set_tos = st->act.set_tos;
5579 			sp->pfs_1400.max_mss = htons(st->act.max_mss);
5580 			sp->pfs_1400.set_prio[0] = st->act.set_prio[0];
5581 			sp->pfs_1400.set_prio[1] = st->act.set_prio[1];
5582 			sp->pfs_1400.rt = st->rt;
5583 			if (st->rt_kif)
5584 				strlcpy(sp->pfs_1400.rt_ifname,
5585 				    st->rt_kif->pfik_name,
5586 				    sizeof(sp->pfs_1400.rt_ifname));
5587 			break;
5588 		default:
5589 			panic("%s: Unsupported pfsync_msg_version %d",
5590 			    __func__, msg_version);
5591 	}
5592 
5593 	if (st->src_node)
5594 		sp->pfs_1301.sync_flags |= PFSYNC_FLAG_SRCNODE;
5595 	if (st->nat_src_node)
5596 		sp->pfs_1301.sync_flags |= PFSYNC_FLAG_NATSRCNODE;
5597 
5598 	sp->pfs_1301.id = st->id;
5599 	sp->pfs_1301.creatorid = st->creatorid;
5600 	pf_state_peer_hton(&st->src, &sp->pfs_1301.src);
5601 	pf_state_peer_hton(&st->dst, &sp->pfs_1301.dst);
5602 
5603 	if (st->rule.ptr == NULL)
5604 		sp->pfs_1301.rule = htonl(-1);
5605 	else
5606 		sp->pfs_1301.rule = htonl(st->rule.ptr->nr);
5607 	if (st->anchor.ptr == NULL)
5608 		sp->pfs_1301.anchor = htonl(-1);
5609 	else
5610 		sp->pfs_1301.anchor = htonl(st->anchor.ptr->nr);
5611 	if (st->nat_rule.ptr == NULL)
5612 		sp->pfs_1301.nat_rule = htonl(-1);
5613 	else
5614 		sp->pfs_1301.nat_rule = htonl(st->nat_rule.ptr->nr);
5615 
5616 	pf_state_counter_hton(st->packets[0], sp->pfs_1301.packets[0]);
5617 	pf_state_counter_hton(st->packets[1], sp->pfs_1301.packets[1]);
5618 	pf_state_counter_hton(st->bytes[0], sp->pfs_1301.bytes[0]);
5619 	pf_state_counter_hton(st->bytes[1], sp->pfs_1301.bytes[1]);
5620 }
5621 
5622 void
5623 pf_state_export(struct pf_state_export *sp, struct pf_kstate *st)
5624 {
5625 	bzero(sp, sizeof(*sp));
5626 
5627 	sp->version = PF_STATE_VERSION;
5628 
5629 	/* copy from state key */
5630 	sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
5631 	sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
5632 	sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
5633 	sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
5634 	sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
5635 	sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
5636 	sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
5637 	sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
5638 	sp->proto = st->key[PF_SK_WIRE]->proto;
5639 	sp->af = st->key[PF_SK_WIRE]->af;
5640 
5641 	/* copy from state */
5642 	strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
5643 	strlcpy(sp->orig_ifname, st->orig_kif->pfik_name,
5644 	    sizeof(sp->orig_ifname));
5645 	bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
5646 	sp->creation = htonl(time_uptime - (st->creation / 1000));
5647 	sp->expire = pf_state_expires(st);
5648 	if (sp->expire <= time_uptime)
5649 		sp->expire = htonl(0);
5650 	else
5651 		sp->expire = htonl(sp->expire - time_uptime);
5652 
5653 	sp->direction = st->direction;
5654 	sp->log = st->act.log;
5655 	sp->timeout = st->timeout;
5656 	/* 8 bits for the old libpfctl, 16 bits for the new libpfctl */
5657 	sp->state_flags_compat = st->state_flags;
5658 	sp->state_flags = htons(st->state_flags);
5659 	if (st->src_node)
5660 		sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
5661 	if (st->nat_src_node)
5662 		sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
5663 
5664 	sp->id = st->id;
5665 	sp->creatorid = st->creatorid;
5666 	pf_state_peer_hton(&st->src, &sp->src);
5667 	pf_state_peer_hton(&st->dst, &sp->dst);
5668 
5669 	if (st->rule.ptr == NULL)
5670 		sp->rule = htonl(-1);
5671 	else
5672 		sp->rule = htonl(st->rule.ptr->nr);
5673 	if (st->anchor.ptr == NULL)
5674 		sp->anchor = htonl(-1);
5675 	else
5676 		sp->anchor = htonl(st->anchor.ptr->nr);
5677 	if (st->nat_rule.ptr == NULL)
5678 		sp->nat_rule = htonl(-1);
5679 	else
5680 		sp->nat_rule = htonl(st->nat_rule.ptr->nr);
5681 
5682 	sp->packets[0] = st->packets[0];
5683 	sp->packets[1] = st->packets[1];
5684 	sp->bytes[0] = st->bytes[0];
5685 	sp->bytes[1] = st->bytes[1];
5686 
5687 	sp->qid = htons(st->act.qid);
5688 	sp->pqid = htons(st->act.pqid);
5689 	sp->dnpipe = htons(st->act.dnpipe);
5690 	sp->dnrpipe = htons(st->act.dnrpipe);
5691 	sp->rtableid = htonl(st->act.rtableid);
5692 	sp->min_ttl = st->act.min_ttl;
5693 	sp->set_tos = st->act.set_tos;
5694 	sp->max_mss = htons(st->act.max_mss);
5695 	sp->rt = st->rt;
5696 	if (st->rt_kif)
5697 		strlcpy(sp->rt_ifname, st->rt_kif->pfik_name,
5698 		    sizeof(sp->rt_ifname));
5699 	sp->set_prio[0] = st->act.set_prio[0];
5700 	sp->set_prio[1] = st->act.set_prio[1];
5701 
5702 }
5703 
5704 static void
5705 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
5706 {
5707 	struct pfr_ktable *kt;
5708 
5709 	KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
5710 
5711 	kt = aw->p.tbl;
5712 	if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
5713 		kt = kt->pfrkt_root;
5714 	aw->p.tbl = NULL;
5715 	aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
5716 		kt->pfrkt_cnt : -1;
5717 }
5718 
5719 static int
5720 pf_add_status_counters(nvlist_t *nvl, const char *name, counter_u64_t *counters,
5721     size_t number, char **names)
5722 {
5723 	nvlist_t        *nvc;
5724 
5725 	nvc = nvlist_create(0);
5726 	if (nvc == NULL)
5727 		return (ENOMEM);
5728 
5729 	for (int i = 0; i < number; i++) {
5730 		nvlist_append_number_array(nvc, "counters",
5731 		    counter_u64_fetch(counters[i]));
5732 		nvlist_append_string_array(nvc, "names",
5733 		    names[i]);
5734 		nvlist_append_number_array(nvc, "ids",
5735 		    i);
5736 	}
5737 	nvlist_add_nvlist(nvl, name, nvc);
5738 	nvlist_destroy(nvc);
5739 
5740 	return (0);
5741 }
5742 
5743 static int
5744 pf_getstatus(struct pfioc_nv *nv)
5745 {
5746 	nvlist_t        *nvl = NULL, *nvc = NULL;
5747 	void            *nvlpacked = NULL;
5748 	int              error;
5749 	struct pf_status s;
5750 	char *pf_reasons[PFRES_MAX+1] = PFRES_NAMES;
5751 	char *pf_lcounter[KLCNT_MAX+1] = KLCNT_NAMES;
5752 	char *pf_fcounter[FCNT_MAX+1] = FCNT_NAMES;
5753 	PF_RULES_RLOCK_TRACKER;
5754 
5755 #define ERROUT(x)      ERROUT_FUNCTION(errout, x)
5756 
5757 	PF_RULES_RLOCK();
5758 
5759 	nvl = nvlist_create(0);
5760 	if (nvl == NULL)
5761 		ERROUT(ENOMEM);
5762 
5763 	nvlist_add_bool(nvl, "running", V_pf_status.running);
5764 	nvlist_add_number(nvl, "since", V_pf_status.since);
5765 	nvlist_add_number(nvl, "debug", V_pf_status.debug);
5766 	nvlist_add_number(nvl, "hostid", V_pf_status.hostid);
5767 	nvlist_add_number(nvl, "states", V_pf_status.states);
5768 	nvlist_add_number(nvl, "src_nodes", V_pf_status.src_nodes);
5769 	nvlist_add_number(nvl, "reass", V_pf_status.reass);
5770 	nvlist_add_bool(nvl, "syncookies_active",
5771 	    V_pf_status.syncookies_active);
5772 	nvlist_add_number(nvl, "halfopen_states", V_pf_status.states_halfopen);
5773 
5774 	/* counters */
5775 	error = pf_add_status_counters(nvl, "counters", V_pf_status.counters,
5776 	    PFRES_MAX, pf_reasons);
5777 	if (error != 0)
5778 		ERROUT(error);
5779 
5780 	/* lcounters */
5781 	error = pf_add_status_counters(nvl, "lcounters", V_pf_status.lcounters,
5782 	    KLCNT_MAX, pf_lcounter);
5783 	if (error != 0)
5784 		ERROUT(error);
5785 
5786 	/* fcounters */
5787 	nvc = nvlist_create(0);
5788 	if (nvc == NULL)
5789 		ERROUT(ENOMEM);
5790 
5791 	for (int i = 0; i < FCNT_MAX; i++) {
5792 		nvlist_append_number_array(nvc, "counters",
5793 		    pf_counter_u64_fetch(&V_pf_status.fcounters[i]));
5794 		nvlist_append_string_array(nvc, "names",
5795 		    pf_fcounter[i]);
5796 		nvlist_append_number_array(nvc, "ids",
5797 		    i);
5798 	}
5799 	nvlist_add_nvlist(nvl, "fcounters", nvc);
5800 	nvlist_destroy(nvc);
5801 	nvc = NULL;
5802 
5803 	/* scounters */
5804 	error = pf_add_status_counters(nvl, "scounters", V_pf_status.scounters,
5805 	    SCNT_MAX, pf_fcounter);
5806 	if (error != 0)
5807 		ERROUT(error);
5808 
5809 	nvlist_add_string(nvl, "ifname", V_pf_status.ifname);
5810 	nvlist_add_binary(nvl, "chksum", V_pf_status.pf_chksum,
5811 	    PF_MD5_DIGEST_LENGTH);
5812 
5813 	pfi_update_status(V_pf_status.ifname, &s);
5814 
5815 	/* pcounters / bcounters */
5816 	for (int i = 0; i < 2; i++) {
5817 		for (int j = 0; j < 2; j++) {
5818 			for (int k = 0; k < 2; k++) {
5819 				nvlist_append_number_array(nvl, "pcounters",
5820 				    s.pcounters[i][j][k]);
5821 			}
5822 			nvlist_append_number_array(nvl, "bcounters",
5823 			    s.bcounters[i][j]);
5824 		}
5825 	}
5826 
5827 	nvlpacked = nvlist_pack(nvl, &nv->len);
5828 	if (nvlpacked == NULL)
5829 		ERROUT(ENOMEM);
5830 
5831 	if (nv->size == 0)
5832 		ERROUT(0);
5833 	else if (nv->size < nv->len)
5834 		ERROUT(ENOSPC);
5835 
5836 	PF_RULES_RUNLOCK();
5837 	error = copyout(nvlpacked, nv->data, nv->len);
5838 	goto done;
5839 
5840 #undef ERROUT
5841 errout:
5842 	PF_RULES_RUNLOCK();
5843 done:
5844 	free(nvlpacked, M_NVLIST);
5845 	nvlist_destroy(nvc);
5846 	nvlist_destroy(nvl);
5847 
5848 	return (error);
5849 }
5850 
5851 /*
5852  * XXX - Check for version mismatch!!!
5853  */
5854 static void
5855 pf_clear_all_states(void)
5856 {
5857 	struct epoch_tracker	 et;
5858 	struct pf_kstate	*s;
5859 	u_int i;
5860 
5861 	NET_EPOCH_ENTER(et);
5862 	for (i = 0; i <= V_pf_hashmask; i++) {
5863 		struct pf_idhash *ih = &V_pf_idhash[i];
5864 relock:
5865 		PF_HASHROW_LOCK(ih);
5866 		LIST_FOREACH(s, &ih->states, entry) {
5867 			s->timeout = PFTM_PURGE;
5868 			/* Don't send out individual delete messages. */
5869 			s->state_flags |= PFSTATE_NOSYNC;
5870 			pf_unlink_state(s);
5871 			goto relock;
5872 		}
5873 		PF_HASHROW_UNLOCK(ih);
5874 	}
5875 	NET_EPOCH_EXIT(et);
5876 }
5877 
5878 static int
5879 pf_clear_tables(void)
5880 {
5881 	struct pfioc_table io;
5882 	int error;
5883 
5884 	bzero(&io, sizeof(io));
5885 	io.pfrio_flags |= PFR_FLAG_ALLRSETS;
5886 
5887 	error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
5888 	    io.pfrio_flags);
5889 
5890 	return (error);
5891 }
5892 
5893 static void
5894 pf_clear_srcnodes(struct pf_ksrc_node *n)
5895 {
5896 	struct pf_kstate *s;
5897 	int i;
5898 
5899 	for (i = 0; i <= V_pf_hashmask; i++) {
5900 		struct pf_idhash *ih = &V_pf_idhash[i];
5901 
5902 		PF_HASHROW_LOCK(ih);
5903 		LIST_FOREACH(s, &ih->states, entry) {
5904 			if (n == NULL || n == s->src_node)
5905 				s->src_node = NULL;
5906 			if (n == NULL || n == s->nat_src_node)
5907 				s->nat_src_node = NULL;
5908 		}
5909 		PF_HASHROW_UNLOCK(ih);
5910 	}
5911 
5912 	if (n == NULL) {
5913 		struct pf_srchash *sh;
5914 
5915 		for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask;
5916 		    i++, sh++) {
5917 			PF_HASHROW_LOCK(sh);
5918 			LIST_FOREACH(n, &sh->nodes, entry) {
5919 				n->expire = 1;
5920 				n->states = 0;
5921 			}
5922 			PF_HASHROW_UNLOCK(sh);
5923 		}
5924 	} else {
5925 		/* XXX: hash slot should already be locked here. */
5926 		n->expire = 1;
5927 		n->states = 0;
5928 	}
5929 }
5930 
5931 static void
5932 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk)
5933 {
5934 	struct pf_ksrc_node_list	 kill;
5935 
5936 	LIST_INIT(&kill);
5937 	for (int i = 0; i <= V_pf_srchashmask; i++) {
5938 		struct pf_srchash *sh = &V_pf_srchash[i];
5939 		struct pf_ksrc_node *sn, *tmp;
5940 
5941 		PF_HASHROW_LOCK(sh);
5942 		LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp)
5943 			if (PF_MATCHA(psnk->psnk_src.neg,
5944 			      &psnk->psnk_src.addr.v.a.addr,
5945 			      &psnk->psnk_src.addr.v.a.mask,
5946 			      &sn->addr, sn->af) &&
5947 			    PF_MATCHA(psnk->psnk_dst.neg,
5948 			      &psnk->psnk_dst.addr.v.a.addr,
5949 			      &psnk->psnk_dst.addr.v.a.mask,
5950 			      &sn->raddr, sn->af)) {
5951 				pf_unlink_src_node(sn);
5952 				LIST_INSERT_HEAD(&kill, sn, entry);
5953 				sn->expire = 1;
5954 			}
5955 		PF_HASHROW_UNLOCK(sh);
5956 	}
5957 
5958 	for (int i = 0; i <= V_pf_hashmask; i++) {
5959 		struct pf_idhash *ih = &V_pf_idhash[i];
5960 		struct pf_kstate *s;
5961 
5962 		PF_HASHROW_LOCK(ih);
5963 		LIST_FOREACH(s, &ih->states, entry) {
5964 			if (s->src_node && s->src_node->expire == 1)
5965 				s->src_node = NULL;
5966 			if (s->nat_src_node && s->nat_src_node->expire == 1)
5967 				s->nat_src_node = NULL;
5968 		}
5969 		PF_HASHROW_UNLOCK(ih);
5970 	}
5971 
5972 	psnk->psnk_killed = pf_free_src_nodes(&kill);
5973 }
5974 
5975 static int
5976 pf_keepcounters(struct pfioc_nv *nv)
5977 {
5978 	nvlist_t	*nvl = NULL;
5979 	void		*nvlpacked = NULL;
5980 	int		 error = 0;
5981 
5982 #define	ERROUT(x)	ERROUT_FUNCTION(on_error, x)
5983 
5984 	if (nv->len > pf_ioctl_maxcount)
5985 		ERROUT(ENOMEM);
5986 
5987 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
5988 	error = copyin(nv->data, nvlpacked, nv->len);
5989 	if (error)
5990 		ERROUT(error);
5991 
5992 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
5993 	if (nvl == NULL)
5994 		ERROUT(EBADMSG);
5995 
5996 	if (! nvlist_exists_bool(nvl, "keep_counters"))
5997 		ERROUT(EBADMSG);
5998 
5999 	V_pf_status.keep_counters = nvlist_get_bool(nvl, "keep_counters");
6000 
6001 on_error:
6002 	nvlist_destroy(nvl);
6003 	free(nvlpacked, M_NVLIST);
6004 	return (error);
6005 }
6006 
6007 unsigned int
6008 pf_clear_states(const struct pf_kstate_kill *kill)
6009 {
6010 	struct pf_state_key_cmp	 match_key;
6011 	struct pf_kstate	*s;
6012 	struct pfi_kkif	*kif;
6013 	int		 idx;
6014 	unsigned int	 killed = 0, dir;
6015 
6016 	NET_EPOCH_ASSERT();
6017 
6018 	for (unsigned int i = 0; i <= V_pf_hashmask; i++) {
6019 		struct pf_idhash *ih = &V_pf_idhash[i];
6020 
6021 relock_DIOCCLRSTATES:
6022 		PF_HASHROW_LOCK(ih);
6023 		LIST_FOREACH(s, &ih->states, entry) {
6024 			/* For floating states look at the original kif. */
6025 			kif = s->kif == V_pfi_all ? s->orig_kif : s->kif;
6026 
6027 			if (kill->psk_ifname[0] &&
6028 			    strcmp(kill->psk_ifname,
6029 			    kif->pfik_name))
6030 				continue;
6031 
6032 			if (kill->psk_kill_match) {
6033 				bzero(&match_key, sizeof(match_key));
6034 
6035 				if (s->direction == PF_OUT) {
6036 					dir = PF_IN;
6037 					idx = PF_SK_STACK;
6038 				} else {
6039 					dir = PF_OUT;
6040 					idx = PF_SK_WIRE;
6041 				}
6042 
6043 				match_key.af = s->key[idx]->af;
6044 				match_key.proto = s->key[idx]->proto;
6045 				PF_ACPY(&match_key.addr[0],
6046 				    &s->key[idx]->addr[1], match_key.af);
6047 				match_key.port[0] = s->key[idx]->port[1];
6048 				PF_ACPY(&match_key.addr[1],
6049 				    &s->key[idx]->addr[0], match_key.af);
6050 				match_key.port[1] = s->key[idx]->port[0];
6051 			}
6052 
6053 			/*
6054 			 * Don't send out individual
6055 			 * delete messages.
6056 			 */
6057 			s->state_flags |= PFSTATE_NOSYNC;
6058 			pf_unlink_state(s);
6059 			killed++;
6060 
6061 			if (kill->psk_kill_match)
6062 				killed += pf_kill_matching_state(&match_key,
6063 				    dir);
6064 
6065 			goto relock_DIOCCLRSTATES;
6066 		}
6067 		PF_HASHROW_UNLOCK(ih);
6068 	}
6069 
6070 	if (V_pfsync_clear_states_ptr != NULL)
6071 		V_pfsync_clear_states_ptr(V_pf_status.hostid, kill->psk_ifname);
6072 
6073 	return (killed);
6074 }
6075 
6076 void
6077 pf_killstates(struct pf_kstate_kill *kill, unsigned int *killed)
6078 {
6079 	struct pf_kstate	*s;
6080 
6081 	NET_EPOCH_ASSERT();
6082 	if (kill->psk_pfcmp.id) {
6083 		if (kill->psk_pfcmp.creatorid == 0)
6084 			kill->psk_pfcmp.creatorid = V_pf_status.hostid;
6085 		if ((s = pf_find_state_byid(kill->psk_pfcmp.id,
6086 		    kill->psk_pfcmp.creatorid))) {
6087 			pf_unlink_state(s);
6088 			*killed = 1;
6089 		}
6090 		return;
6091 	}
6092 
6093 	for (unsigned int i = 0; i <= V_pf_hashmask; i++)
6094 		*killed += pf_killstates_row(kill, &V_pf_idhash[i]);
6095 }
6096 
6097 static int
6098 pf_killstates_nv(struct pfioc_nv *nv)
6099 {
6100 	struct pf_kstate_kill	 kill;
6101 	struct epoch_tracker	 et;
6102 	nvlist_t		*nvl = NULL;
6103 	void			*nvlpacked = NULL;
6104 	int			 error = 0;
6105 	unsigned int		 killed = 0;
6106 
6107 #define ERROUT(x)	ERROUT_FUNCTION(on_error, x)
6108 
6109 	if (nv->len > pf_ioctl_maxcount)
6110 		ERROUT(ENOMEM);
6111 
6112 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6113 	error = copyin(nv->data, nvlpacked, nv->len);
6114 	if (error)
6115 		ERROUT(error);
6116 
6117 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6118 	if (nvl == NULL)
6119 		ERROUT(EBADMSG);
6120 
6121 	error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
6122 	if (error)
6123 		ERROUT(error);
6124 
6125 	NET_EPOCH_ENTER(et);
6126 	pf_killstates(&kill, &killed);
6127 	NET_EPOCH_EXIT(et);
6128 
6129 	free(nvlpacked, M_NVLIST);
6130 	nvlpacked = NULL;
6131 	nvlist_destroy(nvl);
6132 	nvl = nvlist_create(0);
6133 	if (nvl == NULL)
6134 		ERROUT(ENOMEM);
6135 
6136 	nvlist_add_number(nvl, "killed", killed);
6137 
6138 	nvlpacked = nvlist_pack(nvl, &nv->len);
6139 	if (nvlpacked == NULL)
6140 		ERROUT(ENOMEM);
6141 
6142 	if (nv->size == 0)
6143 		ERROUT(0);
6144 	else if (nv->size < nv->len)
6145 		ERROUT(ENOSPC);
6146 
6147 	error = copyout(nvlpacked, nv->data, nv->len);
6148 
6149 on_error:
6150 	nvlist_destroy(nvl);
6151 	free(nvlpacked, M_NVLIST);
6152 	return (error);
6153 }
6154 
6155 static int
6156 pf_clearstates_nv(struct pfioc_nv *nv)
6157 {
6158 	struct pf_kstate_kill	 kill;
6159 	struct epoch_tracker	 et;
6160 	nvlist_t		*nvl = NULL;
6161 	void			*nvlpacked = NULL;
6162 	int			 error = 0;
6163 	unsigned int		 killed;
6164 
6165 #define ERROUT(x)	ERROUT_FUNCTION(on_error, x)
6166 
6167 	if (nv->len > pf_ioctl_maxcount)
6168 		ERROUT(ENOMEM);
6169 
6170 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6171 	error = copyin(nv->data, nvlpacked, nv->len);
6172 	if (error)
6173 		ERROUT(error);
6174 
6175 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6176 	if (nvl == NULL)
6177 		ERROUT(EBADMSG);
6178 
6179 	error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
6180 	if (error)
6181 		ERROUT(error);
6182 
6183 	NET_EPOCH_ENTER(et);
6184 	killed = pf_clear_states(&kill);
6185 	NET_EPOCH_EXIT(et);
6186 
6187 	free(nvlpacked, M_NVLIST);
6188 	nvlpacked = NULL;
6189 	nvlist_destroy(nvl);
6190 	nvl = nvlist_create(0);
6191 	if (nvl == NULL)
6192 		ERROUT(ENOMEM);
6193 
6194 	nvlist_add_number(nvl, "killed", killed);
6195 
6196 	nvlpacked = nvlist_pack(nvl, &nv->len);
6197 	if (nvlpacked == NULL)
6198 		ERROUT(ENOMEM);
6199 
6200 	if (nv->size == 0)
6201 		ERROUT(0);
6202 	else if (nv->size < nv->len)
6203 		ERROUT(ENOSPC);
6204 
6205 	error = copyout(nvlpacked, nv->data, nv->len);
6206 
6207 #undef ERROUT
6208 on_error:
6209 	nvlist_destroy(nvl);
6210 	free(nvlpacked, M_NVLIST);
6211 	return (error);
6212 }
6213 
6214 static int
6215 pf_getstate(struct pfioc_nv *nv)
6216 {
6217 	nvlist_t		*nvl = NULL, *nvls;
6218 	void			*nvlpacked = NULL;
6219 	struct pf_kstate	*s = NULL;
6220 	int			 error = 0;
6221 	uint64_t		 id, creatorid;
6222 
6223 #define ERROUT(x)	ERROUT_FUNCTION(errout, x)
6224 
6225 	if (nv->len > pf_ioctl_maxcount)
6226 		ERROUT(ENOMEM);
6227 
6228 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6229 	error = copyin(nv->data, nvlpacked, nv->len);
6230 	if (error)
6231 		ERROUT(error);
6232 
6233 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6234 	if (nvl == NULL)
6235 		ERROUT(EBADMSG);
6236 
6237 	PFNV_CHK(pf_nvuint64(nvl, "id", &id));
6238 	PFNV_CHK(pf_nvuint64(nvl, "creatorid", &creatorid));
6239 
6240 	s = pf_find_state_byid(id, creatorid);
6241 	if (s == NULL)
6242 		ERROUT(ENOENT);
6243 
6244 	free(nvlpacked, M_NVLIST);
6245 	nvlpacked = NULL;
6246 	nvlist_destroy(nvl);
6247 	nvl = nvlist_create(0);
6248 	if (nvl == NULL)
6249 		ERROUT(ENOMEM);
6250 
6251 	nvls = pf_state_to_nvstate(s);
6252 	if (nvls == NULL)
6253 		ERROUT(ENOMEM);
6254 
6255 	nvlist_add_nvlist(nvl, "state", nvls);
6256 	nvlist_destroy(nvls);
6257 
6258 	nvlpacked = nvlist_pack(nvl, &nv->len);
6259 	if (nvlpacked == NULL)
6260 		ERROUT(ENOMEM);
6261 
6262 	if (nv->size == 0)
6263 		ERROUT(0);
6264 	else if (nv->size < nv->len)
6265 		ERROUT(ENOSPC);
6266 
6267 	error = copyout(nvlpacked, nv->data, nv->len);
6268 
6269 #undef ERROUT
6270 errout:
6271 	if (s != NULL)
6272 		PF_STATE_UNLOCK(s);
6273 	free(nvlpacked, M_NVLIST);
6274 	nvlist_destroy(nvl);
6275 	return (error);
6276 }
6277 
6278 /*
6279  * XXX - Check for version mismatch!!!
6280  */
6281 
6282 /*
6283  * Duplicate pfctl -Fa operation to get rid of as much as we can.
6284  */
6285 static int
6286 shutdown_pf(void)
6287 {
6288 	int error = 0;
6289 	u_int32_t t[5];
6290 	char nn = '\0';
6291 	struct pf_kanchor *anchor;
6292 	struct pf_keth_anchor *eth_anchor;
6293 	int rs_num;
6294 
6295 	do {
6296 		/* Unlink rules of all user defined anchors */
6297 		RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors) {
6298 			/* Wildcard based anchors may not have a respective
6299 			 * explicit anchor rule or they may be left empty
6300 			 * without rules. It leads to anchor.refcnt=0, and the
6301 			 * rest of the logic does not expect it. */
6302 			if (anchor->refcnt == 0)
6303 				anchor->refcnt = 1;
6304 			for (rs_num = 0; rs_num < PF_RULESET_MAX; ++rs_num) {
6305 				if ((error = pf_begin_rules(&t[rs_num], rs_num,
6306 				    anchor->path)) != 0) {
6307 					DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: "
6308 					    "anchor.path=%s rs_num=%d\n",
6309 					    anchor->path, rs_num));
6310 					goto error;	/* XXX: rollback? */
6311 				}
6312 			}
6313 			for (rs_num = 0; rs_num < PF_RULESET_MAX; ++rs_num) {
6314 				error = pf_commit_rules(t[rs_num], rs_num,
6315 				    anchor->path);
6316 				MPASS(error == 0);
6317 			}
6318 		}
6319 
6320 		/* Unlink rules of all user defined ether anchors */
6321 		RB_FOREACH(eth_anchor, pf_keth_anchor_global,
6322 		    &V_pf_keth_anchors) {
6323 			/* Wildcard based anchors may not have a respective
6324 			 * explicit anchor rule or they may be left empty
6325 			 * without rules. It leads to anchor.refcnt=0, and the
6326 			 * rest of the logic does not expect it. */
6327 			if (eth_anchor->refcnt == 0)
6328 				eth_anchor->refcnt = 1;
6329 			if ((error = pf_begin_eth(&t[0], eth_anchor->path))
6330 			    != 0) {
6331 				DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: eth "
6332 				    "anchor.path=%s\n", eth_anchor->path));
6333 				goto error;
6334 			}
6335 			error = pf_commit_eth(t[0], eth_anchor->path);
6336 			MPASS(error == 0);
6337 		}
6338 
6339 		if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
6340 		    != 0) {
6341 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
6342 			break;
6343 		}
6344 		if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
6345 		    != 0) {
6346 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
6347 			break;		/* XXX: rollback? */
6348 		}
6349 		if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
6350 		    != 0) {
6351 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
6352 			break;		/* XXX: rollback? */
6353 		}
6354 		if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
6355 		    != 0) {
6356 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
6357 			break;		/* XXX: rollback? */
6358 		}
6359 		if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
6360 		    != 0) {
6361 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
6362 			break;		/* XXX: rollback? */
6363 		}
6364 
6365 		error = pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
6366 		MPASS(error == 0);
6367 		error = pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
6368 		MPASS(error == 0);
6369 		error = pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
6370 		MPASS(error == 0);
6371 		error = pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
6372 		MPASS(error == 0);
6373 		error = pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
6374 		MPASS(error == 0);
6375 
6376 		if ((error = pf_clear_tables()) != 0)
6377 			break;
6378 
6379 		if ((error = pf_begin_eth(&t[0], &nn)) != 0) {
6380 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: eth\n"));
6381 			break;
6382 		}
6383 		error = pf_commit_eth(t[0], &nn);
6384 		MPASS(error == 0);
6385 
6386 #ifdef ALTQ
6387 		if ((error = pf_begin_altq(&t[0])) != 0) {
6388 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
6389 			break;
6390 		}
6391 		pf_commit_altq(t[0]);
6392 #endif
6393 
6394 		pf_clear_all_states();
6395 
6396 		pf_clear_srcnodes(NULL);
6397 
6398 		/* status does not use malloced mem so no need to cleanup */
6399 		/* fingerprints and interfaces have their own cleanup code */
6400 	} while(0);
6401 
6402 error:
6403 	return (error);
6404 }
6405 
6406 static pfil_return_t
6407 pf_check_return(int chk, struct mbuf **m)
6408 {
6409 
6410 	switch (chk) {
6411 	case PF_PASS:
6412 		if (*m == NULL)
6413 			return (PFIL_CONSUMED);
6414 		else
6415 			return (PFIL_PASS);
6416 		break;
6417 	default:
6418 		if (*m != NULL) {
6419 			m_freem(*m);
6420 			*m = NULL;
6421 		}
6422 		return (PFIL_DROPPED);
6423 	}
6424 }
6425 
6426 static pfil_return_t
6427 pf_eth_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
6428     void *ruleset __unused, struct inpcb *inp)
6429 {
6430 	int chk;
6431 
6432 	chk = pf_test_eth(PF_IN, flags, ifp, m, inp);
6433 
6434 	return (pf_check_return(chk, m));
6435 }
6436 
6437 static pfil_return_t
6438 pf_eth_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
6439     void *ruleset __unused, struct inpcb *inp)
6440 {
6441 	int chk;
6442 
6443 	chk = pf_test_eth(PF_OUT, flags, ifp, m, inp);
6444 
6445 	return (pf_check_return(chk, m));
6446 }
6447 
6448 #ifdef INET
6449 static pfil_return_t
6450 pf_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
6451     void *ruleset __unused, struct inpcb *inp)
6452 {
6453 	int chk;
6454 
6455 	chk = pf_test(PF_IN, flags, ifp, m, inp, NULL);
6456 
6457 	return (pf_check_return(chk, m));
6458 }
6459 
6460 static pfil_return_t
6461 pf_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
6462     void *ruleset __unused,  struct inpcb *inp)
6463 {
6464 	int chk;
6465 
6466 	chk = pf_test(PF_OUT, flags, ifp, m, inp, NULL);
6467 
6468 	return (pf_check_return(chk, m));
6469 }
6470 #endif
6471 
6472 #ifdef INET6
6473 static pfil_return_t
6474 pf_check6_in(struct mbuf **m, struct ifnet *ifp, int flags,
6475     void *ruleset __unused,  struct inpcb *inp)
6476 {
6477 	int chk;
6478 
6479 	/*
6480 	 * In case of loopback traffic IPv6 uses the real interface in
6481 	 * order to support scoped addresses. In order to support stateful
6482 	 * filtering we have change this to lo0 as it is the case in IPv4.
6483 	 */
6484 	CURVNET_SET(ifp->if_vnet);
6485 	chk = pf_test6(PF_IN, flags, (*m)->m_flags & M_LOOP ? V_loif : ifp,
6486 	    m, inp, NULL);
6487 	CURVNET_RESTORE();
6488 
6489 	return (pf_check_return(chk, m));
6490 }
6491 
6492 static pfil_return_t
6493 pf_check6_out(struct mbuf **m, struct ifnet *ifp, int flags,
6494     void *ruleset __unused,  struct inpcb *inp)
6495 {
6496 	int chk;
6497 
6498 	CURVNET_SET(ifp->if_vnet);
6499 	chk = pf_test6(PF_OUT, flags, ifp, m, inp, NULL);
6500 	CURVNET_RESTORE();
6501 
6502 	return (pf_check_return(chk, m));
6503 }
6504 #endif /* INET6 */
6505 
6506 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_in_hook);
6507 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_out_hook);
6508 #define	V_pf_eth_in_hook	VNET(pf_eth_in_hook)
6509 #define	V_pf_eth_out_hook	VNET(pf_eth_out_hook)
6510 
6511 #ifdef INET
6512 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_in_hook);
6513 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_out_hook);
6514 #define	V_pf_ip4_in_hook	VNET(pf_ip4_in_hook)
6515 #define	V_pf_ip4_out_hook	VNET(pf_ip4_out_hook)
6516 #endif
6517 #ifdef INET6
6518 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_in_hook);
6519 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_out_hook);
6520 #define	V_pf_ip6_in_hook	VNET(pf_ip6_in_hook)
6521 #define	V_pf_ip6_out_hook	VNET(pf_ip6_out_hook)
6522 #endif
6523 
6524 static void
6525 hook_pf_eth(void)
6526 {
6527 	struct pfil_hook_args pha = {
6528 		.pa_version = PFIL_VERSION,
6529 		.pa_modname = "pf",
6530 		.pa_type = PFIL_TYPE_ETHERNET,
6531 	};
6532 	struct pfil_link_args pla = {
6533 		.pa_version = PFIL_VERSION,
6534 	};
6535 	int ret __diagused;
6536 
6537 	if (atomic_load_bool(&V_pf_pfil_eth_hooked))
6538 		return;
6539 
6540 	pha.pa_mbuf_chk = pf_eth_check_in;
6541 	pha.pa_flags = PFIL_IN;
6542 	pha.pa_rulname = "eth-in";
6543 	V_pf_eth_in_hook = pfil_add_hook(&pha);
6544 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6545 	pla.pa_head = V_link_pfil_head;
6546 	pla.pa_hook = V_pf_eth_in_hook;
6547 	ret = pfil_link(&pla);
6548 	MPASS(ret == 0);
6549 	pha.pa_mbuf_chk = pf_eth_check_out;
6550 	pha.pa_flags = PFIL_OUT;
6551 	pha.pa_rulname = "eth-out";
6552 	V_pf_eth_out_hook = pfil_add_hook(&pha);
6553 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6554 	pla.pa_head = V_link_pfil_head;
6555 	pla.pa_hook = V_pf_eth_out_hook;
6556 	ret = pfil_link(&pla);
6557 	MPASS(ret == 0);
6558 
6559 	atomic_store_bool(&V_pf_pfil_eth_hooked, true);
6560 }
6561 
6562 static void
6563 hook_pf(void)
6564 {
6565 	struct pfil_hook_args pha = {
6566 		.pa_version = PFIL_VERSION,
6567 		.pa_modname = "pf",
6568 	};
6569 	struct pfil_link_args pla = {
6570 		.pa_version = PFIL_VERSION,
6571 	};
6572 	int ret __diagused;
6573 
6574 	if (atomic_load_bool(&V_pf_pfil_hooked))
6575 		return;
6576 
6577 #ifdef INET
6578 	pha.pa_type = PFIL_TYPE_IP4;
6579 	pha.pa_mbuf_chk = pf_check_in;
6580 	pha.pa_flags = PFIL_IN;
6581 	pha.pa_rulname = "default-in";
6582 	V_pf_ip4_in_hook = pfil_add_hook(&pha);
6583 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6584 	pla.pa_head = V_inet_pfil_head;
6585 	pla.pa_hook = V_pf_ip4_in_hook;
6586 	ret = pfil_link(&pla);
6587 	MPASS(ret == 0);
6588 	pha.pa_mbuf_chk = pf_check_out;
6589 	pha.pa_flags = PFIL_OUT;
6590 	pha.pa_rulname = "default-out";
6591 	V_pf_ip4_out_hook = pfil_add_hook(&pha);
6592 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6593 	pla.pa_head = V_inet_pfil_head;
6594 	pla.pa_hook = V_pf_ip4_out_hook;
6595 	ret = pfil_link(&pla);
6596 	MPASS(ret == 0);
6597 	if (V_pf_filter_local) {
6598 		pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6599 		pla.pa_head = V_inet_local_pfil_head;
6600 		pla.pa_hook = V_pf_ip4_out_hook;
6601 		ret = pfil_link(&pla);
6602 		MPASS(ret == 0);
6603 	}
6604 #endif
6605 #ifdef INET6
6606 	pha.pa_type = PFIL_TYPE_IP6;
6607 	pha.pa_mbuf_chk = pf_check6_in;
6608 	pha.pa_flags = PFIL_IN;
6609 	pha.pa_rulname = "default-in6";
6610 	V_pf_ip6_in_hook = pfil_add_hook(&pha);
6611 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6612 	pla.pa_head = V_inet6_pfil_head;
6613 	pla.pa_hook = V_pf_ip6_in_hook;
6614 	ret = pfil_link(&pla);
6615 	MPASS(ret == 0);
6616 	pha.pa_mbuf_chk = pf_check6_out;
6617 	pha.pa_rulname = "default-out6";
6618 	pha.pa_flags = PFIL_OUT;
6619 	V_pf_ip6_out_hook = pfil_add_hook(&pha);
6620 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6621 	pla.pa_head = V_inet6_pfil_head;
6622 	pla.pa_hook = V_pf_ip6_out_hook;
6623 	ret = pfil_link(&pla);
6624 	MPASS(ret == 0);
6625 	if (V_pf_filter_local) {
6626 		pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6627 		pla.pa_head = V_inet6_local_pfil_head;
6628 		pla.pa_hook = V_pf_ip6_out_hook;
6629 		ret = pfil_link(&pla);
6630 		MPASS(ret == 0);
6631 	}
6632 #endif
6633 
6634 	atomic_store_bool(&V_pf_pfil_hooked, true);
6635 }
6636 
6637 static void
6638 dehook_pf_eth(void)
6639 {
6640 
6641 	if (!atomic_load_bool(&V_pf_pfil_eth_hooked))
6642 		return;
6643 
6644 	pfil_remove_hook(V_pf_eth_in_hook);
6645 	pfil_remove_hook(V_pf_eth_out_hook);
6646 
6647 	atomic_store_bool(&V_pf_pfil_eth_hooked, false);
6648 }
6649 
6650 static void
6651 dehook_pf(void)
6652 {
6653 
6654 	if (!atomic_load_bool(&V_pf_pfil_hooked))
6655 		return;
6656 
6657 #ifdef INET
6658 	pfil_remove_hook(V_pf_ip4_in_hook);
6659 	pfil_remove_hook(V_pf_ip4_out_hook);
6660 #endif
6661 #ifdef INET6
6662 	pfil_remove_hook(V_pf_ip6_in_hook);
6663 	pfil_remove_hook(V_pf_ip6_out_hook);
6664 #endif
6665 
6666 	atomic_store_bool(&V_pf_pfil_hooked, false);
6667 }
6668 
6669 static void
6670 pf_load_vnet(void)
6671 {
6672 	V_pf_tag_z = uma_zcreate("pf tags", sizeof(struct pf_tagname),
6673 	    NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
6674 
6675 	rm_init_flags(&V_pf_rules_lock, "pf rulesets", RM_RECURSE);
6676 	sx_init(&V_pf_ioctl_lock, "pf ioctl");
6677 
6678 	pf_init_tagset(&V_pf_tags, &pf_rule_tag_hashsize,
6679 	    PF_RULE_TAG_HASH_SIZE_DEFAULT);
6680 #ifdef ALTQ
6681 	pf_init_tagset(&V_pf_qids, &pf_queue_tag_hashsize,
6682 	    PF_QUEUE_TAG_HASH_SIZE_DEFAULT);
6683 #endif
6684 
6685 	V_pf_keth = &V_pf_main_keth_anchor.ruleset;
6686 
6687 	pfattach_vnet();
6688 	V_pf_vnet_active = 1;
6689 }
6690 
6691 static int
6692 pf_load(void)
6693 {
6694 	int error;
6695 
6696 	sx_init(&pf_end_lock, "pf end thread");
6697 
6698 	pf_mtag_initialize();
6699 
6700 	pf_dev = make_dev(&pf_cdevsw, 0, UID_ROOT, GID_WHEEL, 0600, PF_NAME);
6701 	if (pf_dev == NULL)
6702 		return (ENOMEM);
6703 
6704 	pf_end_threads = 0;
6705 	error = kproc_create(pf_purge_thread, NULL, &pf_purge_proc, 0, 0, "pf purge");
6706 	if (error != 0)
6707 		return (error);
6708 
6709 	pfi_initialize();
6710 
6711 	return (0);
6712 }
6713 
6714 static void
6715 pf_unload_vnet(void)
6716 {
6717 	int ret __diagused;
6718 
6719 	V_pf_vnet_active = 0;
6720 	V_pf_status.running = 0;
6721 	dehook_pf();
6722 	dehook_pf_eth();
6723 
6724 	PF_RULES_WLOCK();
6725 	pf_syncookies_cleanup();
6726 	shutdown_pf();
6727 	PF_RULES_WUNLOCK();
6728 
6729 	/* Make sure we've cleaned up ethernet rules before we continue. */
6730 	NET_EPOCH_DRAIN_CALLBACKS();
6731 
6732 	ret = swi_remove(V_pf_swi_cookie);
6733 	MPASS(ret == 0);
6734 	ret = intr_event_destroy(V_pf_swi_ie);
6735 	MPASS(ret == 0);
6736 
6737 	pf_unload_vnet_purge();
6738 
6739 	pf_normalize_cleanup();
6740 	PF_RULES_WLOCK();
6741 	pfi_cleanup_vnet();
6742 	PF_RULES_WUNLOCK();
6743 	pfr_cleanup();
6744 	pf_osfp_flush();
6745 	pf_cleanup();
6746 	if (IS_DEFAULT_VNET(curvnet))
6747 		pf_mtag_cleanup();
6748 
6749 	pf_cleanup_tagset(&V_pf_tags);
6750 #ifdef ALTQ
6751 	pf_cleanup_tagset(&V_pf_qids);
6752 #endif
6753 	uma_zdestroy(V_pf_tag_z);
6754 
6755 #ifdef PF_WANT_32_TO_64_COUNTER
6756 	PF_RULES_WLOCK();
6757 	LIST_REMOVE(V_pf_kifmarker, pfik_allkiflist);
6758 
6759 	MPASS(LIST_EMPTY(&V_pf_allkiflist));
6760 	MPASS(V_pf_allkifcount == 0);
6761 
6762 	LIST_REMOVE(&V_pf_default_rule, allrulelist);
6763 	V_pf_allrulecount--;
6764 	LIST_REMOVE(V_pf_rulemarker, allrulelist);
6765 
6766 	MPASS(LIST_EMPTY(&V_pf_allrulelist));
6767 	MPASS(V_pf_allrulecount == 0);
6768 
6769 	PF_RULES_WUNLOCK();
6770 
6771 	free(V_pf_kifmarker, PFI_MTYPE);
6772 	free(V_pf_rulemarker, M_PFRULE);
6773 #endif
6774 
6775 	/* Free counters last as we updated them during shutdown. */
6776 	pf_counter_u64_deinit(&V_pf_default_rule.evaluations);
6777 	for (int i = 0; i < 2; i++) {
6778 		pf_counter_u64_deinit(&V_pf_default_rule.packets[i]);
6779 		pf_counter_u64_deinit(&V_pf_default_rule.bytes[i]);
6780 	}
6781 	counter_u64_free(V_pf_default_rule.states_cur);
6782 	counter_u64_free(V_pf_default_rule.states_tot);
6783 	counter_u64_free(V_pf_default_rule.src_nodes);
6784 	uma_zfree_pcpu(pf_timestamp_pcpu_zone, V_pf_default_rule.timestamp);
6785 
6786 	for (int i = 0; i < PFRES_MAX; i++)
6787 		counter_u64_free(V_pf_status.counters[i]);
6788 	for (int i = 0; i < KLCNT_MAX; i++)
6789 		counter_u64_free(V_pf_status.lcounters[i]);
6790 	for (int i = 0; i < FCNT_MAX; i++)
6791 		pf_counter_u64_deinit(&V_pf_status.fcounters[i]);
6792 	for (int i = 0; i < SCNT_MAX; i++)
6793 		counter_u64_free(V_pf_status.scounters[i]);
6794 
6795 	rm_destroy(&V_pf_rules_lock);
6796 	sx_destroy(&V_pf_ioctl_lock);
6797 }
6798 
6799 static void
6800 pf_unload(void)
6801 {
6802 
6803 	sx_xlock(&pf_end_lock);
6804 	pf_end_threads = 1;
6805 	while (pf_end_threads < 2) {
6806 		wakeup_one(pf_purge_thread);
6807 		sx_sleep(pf_purge_proc, &pf_end_lock, 0, "pftmo", 0);
6808 	}
6809 	sx_xunlock(&pf_end_lock);
6810 
6811 	pf_nl_unregister();
6812 
6813 	if (pf_dev != NULL)
6814 		destroy_dev(pf_dev);
6815 
6816 	pfi_cleanup();
6817 
6818 	sx_destroy(&pf_end_lock);
6819 }
6820 
6821 static void
6822 vnet_pf_init(void *unused __unused)
6823 {
6824 
6825 	pf_load_vnet();
6826 }
6827 VNET_SYSINIT(vnet_pf_init, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
6828     vnet_pf_init, NULL);
6829 
6830 static void
6831 vnet_pf_uninit(const void *unused __unused)
6832 {
6833 
6834 	pf_unload_vnet();
6835 }
6836 SYSUNINIT(pf_unload, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND, pf_unload, NULL);
6837 VNET_SYSUNINIT(vnet_pf_uninit, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
6838     vnet_pf_uninit, NULL);
6839 
6840 static int
6841 pf_modevent(module_t mod, int type, void *data)
6842 {
6843 	int error = 0;
6844 
6845 	switch(type) {
6846 	case MOD_LOAD:
6847 		error = pf_load();
6848 		pf_nl_register();
6849 		break;
6850 	case MOD_UNLOAD:
6851 		/* Handled in SYSUNINIT(pf_unload) to ensure it's done after
6852 		 * the vnet_pf_uninit()s */
6853 		break;
6854 	default:
6855 		error = EINVAL;
6856 		break;
6857 	}
6858 
6859 	return (error);
6860 }
6861 
6862 static moduledata_t pf_mod = {
6863 	"pf",
6864 	pf_modevent,
6865 	0
6866 };
6867 
6868 DECLARE_MODULE(pf, pf_mod, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND);
6869 MODULE_DEPEND(pf, netlink, 1, 1, 1);
6870 MODULE_VERSION(pf, PF_MODVER);
6871